Decoding Evolutionary Signals: A Practical Guide to Distinguishing Incomplete Lineage Sorting from Introgression in Phylogenomic Data

Abstract

Accurately distinguishing between incomplete lineage sorting (ILS) and introgression is a critical challenge in phylogenomics, with profound implications for understanding evolutionary history, species delimitation, and biomedical applications such as drug target identification and pathogen evolution tracking. This article provides a comprehensive framework for researchers and drug development professionals, covering foundational concepts, state-of-the-art methodological approaches, troubleshooting strategies for complex datasets, and validation techniques. By synthesizing current literature and practical case studies, we offer a definitive guide for navigating gene tree discordance to reveal true evolutionary histories, ultimately enhancing the reliability of phylogenetic inferences in basic research and therapeutic development.

The Genomic Discordance Problem: Understanding ILS and Introgression

Troubleshooting Guides

Guide 1: Diagnosing the Primary Source of Gene Tree Discordance

Q: I have observed widespread incongruence among my gene trees. How can I determine if it is caused by Incomplete Lineage Sorting (ILS) or introgression?

A: Disentangling these sources requires a combination of phylogenetic and population genetic approaches. The table below outlines the key diagnostic patterns.

Table 1: Diagnostic Patterns for ILS vs. Introgression

Feature	Incomplete Lineage Sorting (ILS)	Introgression/Hybridization
Expected Gene Tree Frequencies	The two discordant gene tree topologies are expected to be equal in frequency [1].	The two discordant gene tree topologies are expected to be imbalanced, with one discordant topology over-represented [1].
Phylogenetic Signal	Can cause cytoplasmic-nuclear discordance, but organelle genomes typically share a common history.	Often leads to strong conflict between cytoplasmic (e.g., chloroplast, mitochondrial) and nuclear phylogenies [2].
Genomic Landscape	Discordance is relatively uniform across the genome.	Creates a heterogeneous landscape; introgressed regions are clustered in "blocks" with reduced discordance in between [1].
Useful Detection Methods	Multi-species coalescent (MSC) model; site concordance factors (sCF).	D-statistics (ABBA-BABA); Phylogenetic Networks; QuIBL [3] [1].

Experimental Protocol: A Step-by-Step Workflow for Diagnosis

• Infer Gene Trees and Species Tree: Estimate gene trees from numerous, independent loci (e.g., 1,000+ nuclear orthologous genes). Reconstruct a species tree using both concatenation and coalescent-based methods (e.g., ASTRAL) [3] [2].

• Quantify Discordance: Calculate gene tree frequencies and use metrics like "site concordance factors" (sCF) to identify nodes with high disagreement [3].

• Apply the D-statistic Test: This test uses patterns of allele sharing (e.g., ABBA-BABA patterns) among four taxa to detect a significant excess of shared derived alleles between non-sister species, which is a signature of introgression [1] [4].

• Test for Imbalanced Gene Trees: For nodes with high discordance, check if the frequencies of the two discordant topologies are significantly different. Imbalance suggests introgression, while equal frequencies are consistent with ILS [1].

• Reconstruct Phylogenetic Networks: If introgression is suggested, use network-based methods (e.g., PhyloNet) to model hybridization events explicitly [5].

Q: My data suggests both ILS and introgression are present, and their signals are混淆. How can I quantify their relative contributions?

A: Many evolutionary histories involve a mixture of processes. A recent study on Fagaceae provides a framework for decomposition analysis [2].

Experimental Protocol: Decomposition Analysis

• Identify a Robust Species Tree: Use a method that accounts for ILS as a baseline species history.
• Categorize Gene Trees: Classify each gene tree based on its topology relative to the species tree (concordant vs. discordant).
• Model Alternative Processes:
  • Gene Tree Estimation Error (GTEE): Estimate the error rate of your gene tree reconstruction, for example, via bootstrap support [2].
  • ILS Expectation: Under the pure ILS model, calculate the expected frequency of the major discordant topology.
  • Introgression Signal: The excess of the major discordant topology beyond the ILS expectation, after accounting for GTEE, can be attributed to introgression.

Table 2: Example Contribution Breakdown from a Phylogenomic Study [2]

Source of Discordance	Contribution to Gene Tree Variation
Gene Tree Estimation Error (GTEE)	21.19%
Incomplete Lineage Sorting (ILS)	9.84%
Gene Flow (Introgression)	7.76%
Total Accounted Discordance	38.79%

Frequently Asked Questions (FAQs)

Q1: Can I use organelle genomes (e.g., chloroplast or mitochondrial DNA) to distinguish ILS from introgression?

A: Yes. Since organelle genomes are often uniparentally inherited and do not recombine in the same way as nuclear genes, they have different histories. A strong, well-supported conflict between a cytoplasmic genome tree and the nuclear species tree is a classic signature of historical introgression (often chloroplast capture in plants) [2]. ILS can also cause discordance, but the specific pattern is key.

Q2: What are the minimum data requirements for testing for introgression?

A: The minimum requirement is genomic data from a single individual from each of three focal species and an outgroup (a rooted triplet or unrooted quartet) [1]. This data structure allows for powerful tests like the D-statistic. However, more comprehensive sampling within species provides greater power and robustness.

Q3: My study system underwent a rapid radiation. What is the biggest challenge in resolving its phylogeny?

A: Rapid radiations are characterized by short internal branches on the species tree. This directly increases the probability of ILS because ancestral polymorphisms have little time to coalesce. It also provides a narrow window for hybridization, making the signals of ILS and introgression particularly difficult to disentangle, as seen in groups like Fagaceae [2] and Tulipeae [3]. In such cases, a combination of many loci and methods that account for both processes is essential.

Q4: Can natural selection mislead my analysis?

A: Yes. Natural selection, particularly convergent evolution, can cause non-vertical inheritance of genetic signals. For instance, if genes under positive selection for the same trait in different lineages are included, they may group together based on convergent adaptation rather than shared ancestry, creating a misleading phylogenetic signal [4]. Filtering datasets or conducting separate analyses on different functional gene sets can help mitigate this.

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Solutions for Phylogenomic Analysis of Discordance

Research Reagent / Tool	Function / Application
Transcriptome Sequencing	Provides thousands of low-copy nuclear orthologous genes for robust phylogenomic analysis without the need for a reference genome [3] [4].
D-statistic (ABBA-BABA)	A summary statistic-based method used as a primary test to detect significant introgression against an ILS null hypothesis [1] [4].
Multi-Species Coalescent (MSC) Model	A probabilistic framework that models ILS explicitly, used in species tree inference (e.g., ASTRAL) and as a null model for introgression tests [3] [1].
Phylogenetic Network Inference	Model-based methods (e.g., in PhyloNet) that represent evolutionary history as a network, simultaneously accounting for both ILS and hybridization/introgression [5].
Site Concordance Factors (sCF)	Measures the percentage of decisive alignment sites supporting a given branch in a species tree, helping to identify nodes with pervasive discordance [3].

Troubleshooting Guide: Distinguishing ILS from Introgression

Frequently Asked Questions

Q1: My gene trees are highly discordant. How can I determine if the cause is Incomplete Lineage Sorting (ILS) or introgression? Discordant gene trees can stem from ILS, introgression, or both. Key steps to distinguish them include:

• Conduct Quartet-based Tests: Use statistical methods like the D-statistic (ABBA-BABA test) to detect signatures of introgression. A significant D-statistic indicates gene flow between taxa, rejecting the null hypothesis of ILS-only evolution [6] [7].
• Analyze Site Patterns: Calculate site concordance factors (sCF) and site discordance factors (sDF). An sCF near 33% at a specific node strongly suggests that ILS is the predominant cause of discordance, as this is the expected value under a random coalescent process. Imbalanced sDF1/sDF2 values can point towards introgression [7].
• Perform Polytomy Tests: Compare the likelihood of a bifurcating tree versus a polytomy. If a multifurcating tree is not significantly less likely, it supports a rapid radiation scenario where ILS is expected to be extensive [7].
• Use Phylogenetic Network Analysis: When specific nodes show high or imbalanced discordance, phylogenetic network models (e.g., using PhyloNet or SplitsTree) can visualize and test for potential reticulate evolutionary events like introgression [7].

Q2: What genomic features can serve as "smoking guns" for Horizontal Gene Transfer (HGT) versus vertical descent? HGT events often leave distinct genomic signatures that differ from vertical inheritance and ILS.

• Unexpected Phylogenetic Affiliation: The most direct evidence comes from phylogenies where a gene from the recipient species clusters with orthologs from a distantly related donor species (e.g., a plant gene clustering with bacterial sequences) with strong statistical support, contradicting the established species tree [8] [9] [10].
• Anomalous Nucleotide Composition: Recently transferred genes may retain the nucleotide bias (e.g., GC content) and codon usage preferences of the donor genome, making them stand out from the rest of the recipient's genome. Note that this signal erodes over time [10].
• Presence of Mobile Genetic Elements: The proximity of a gene to transposons, phage integrase sites, or plasmid-related sequences can provide mechanistic evidence for its mobility and integration into the host genome [10].
• Patchy Distribution: A gene found in one species or lineage but absent from its close relatives, yet present in a distantly related lineage, is a classic indicator of HGT [9] [10].

Q3: In a rapid radiation, why is ILS so pervasive, and how does it confound species tree reconstruction? During rapid speciation events, insufficient time passes for ancestral genetic polymorphisms to become fixed in the new daughter lineages. This means that multiple divergent alleles of a single gene can be passed down through the speciation events, leading to gene trees that reflect the history of the allele rather than the species [6]. With thousands of genes, this results in a high percentage of gene trees being discordant with the overall species tree. Standard concatenation methods can be misled by this widespread discordance, inferring an incorrect species tree. Coalescent-based methods, which model this process explicitly, are required for more accurate reconstruction [7].

Q4: Are there specific genomic markers that are more reliable for distinguishing ILS from introgression? Yes, different markers have different properties:

• SINEs and other Retrotransposons: The presence/absence patterns of SINE insertions are considered nearly perfect phylogenetic markers because they are homoplasy-free (rarely undergo parallel insertions or precise excisions). Widespread discordance in SINE patterns is, therefore, strong evidence for ILS, while a few specific discordant insertions can indicate introgression [6].
• Multi-locus Nuclear Data: Using hundreds to thousands of low-copy nuclear orthologous genes provides the statistical power to separate the signal (species tree) from the noise (ILS and introgression) through coalescent analysis [7].
• Organellar Genomes (Plastid/Mitochondrial): These are typically uniparentially inherited and can have different evolutionary histories from the nuclear genome. Incongruence between organellar and nuclear trees can be a clear sign of past introgression or hybridization (e.g., chloroplast capture) [7].

Key Experimental Protocols

Protocol 1: D-Statistic (ABBA-BABA Test) for Introgression Detection

Purpose: To test for gene flow between a closely related "test" taxon and a more distantly related "sister" taxon, which would violate the expected evolutionary tree.

Methodology:

• Taxon Configuration: Define four taxa in a rooted phylogeny: P1, P2, P3, and an outgroup (O). The hypothesis is that P2 and P3 are sister species, but there might be gene flow between P3 and P1.
• Site Pattern Counting: Scan the aligned genomic sequences (e.g., from whole genomes, UCEs, or transcriptomes) for bi-allelic sites and count the occurrences of two specific patterns:
  • ABBA Sites: The outgroup (O) has allele A, P1 has allele B, P2 has allele B, and P3 has allele A.
  • BABA Sites: The outgroup (O) has allele A, P1 has allele B, P2 has allele A, and P3 has allele B.
• Calculation and Significance Testing: Calculate the D-statistic as D = (ABBA - BABA) / (ABBA + BABA). Under no introgression, ABBA and BABA are equally likely (D ≈ 0). A significant positive or negative D-value indicates an excess of ABBA or BABA sites, respectively, suggesting introgression. Significance is assessed using a block jackknife or Z-score test [6] [7].

Workflow Visualization:

Protocol 2: Phylogenomic Analysis for ILS Assessment

Purpose: To reconstruct a robust species tree in the presence of widespread gene tree discordance and to quantify the contribution of ILS.

Methodology:

• Data Collection: Sequence transcriptomes or use hybrid-capture methods (e.g., for UCEs) to generate data for hundreds to thousands of nuclear orthologous genes across multiple taxa [7].
• Gene Tree Inference: Infer individual maximum likelihood gene trees for each orthologous locus.
• Species Tree Inference: Reconstruct the species tree using both concatenation (e.g., IQ-TREE) and multi-species coalescent methods (e.g., ASTRAL-III). A large discrepancy between the two trees suggests substantial ILS [7].
• Quantify Discordance: Calculate gene concordance factors (gCF) and site concordance factors (sCF) for each node in the species tree. A low gCF/sCF (approaching 33%) indicates that the node is poorly supported by the data due to processes like ILS [7].

Workflow Visualization:

Comparative Data Tables

Table 1: Diagnostic Features of ILS vs. Introgression

Feature	Incomplete Lineage Sorting (ILS)	Introgression / Hybridization
Underlying Cause	Retention of ancestral genetic variation due to rapid speciation [6] [7]	Transfer of genetic material between two divergent lineages [8] [10]
Phylogenetic Signal	Randomly distributed discordance across the genome; all possible gene tree topologies are represented [6] [7]	Directional discordance; gene trees consistently group the introgressing taxa [6]
Expected Site Patterns (D-statistic)	D ≈ 0 (No significant excess of ABBA or BABA sites) [6]	D significantly different from 0 (Excess of ABBA or BABA sites) [6] [7]
Concordance Factors	Site Concordance Factor (sCF) for a node is expected to be ~33% [7]	Site Concordance Factor (sCF) is not necessarily expected to be 33%
Genomic Blockiness	Discordant signals are not clustered in specific genomic regions	Discordant signals can be clustered in specific genomic blocks (haplotypes) inherited from the donor species

Table 2: Documented Cases and Functional Impacts of HGT in Plants

This table summarizes quantitative data on horizontal gene transfer events from the scientific literature [8].

Transfer Type	Donor	Recipient	Number of HGTs	Functional Impact
Plant-Plant	Various Grass Species	Alloteropsis semialata	Hundreds among grasses [8]	Stress response, photosynthetic efficiency (C4 pathway) [8]
Plant-Plant	Various Hosts	Parasitic Plants (Cuscuta, Striga)	Hundreds (42% of reported plant-plant HGTs) [8]	Enhanced parasitic ability, haustorium development [8]
Plant-Prokaryote	Bacteria	Triticeae (wheat, barley)	Not specified	Enhanced drought tolerance, improved photosynthesis [8]
Plant-Prokaryote	Bacteria	Ferns (Azolla)	Not specified	High insect resistance [8]
Plant-Fungi	Fungi	Cycas panzhihuaensis	Not specified	Production of an insecticidal toxin [8]

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Reagents for Phylogenomic Conflict Analysis

Item	Function/Brief Explanation
Ultra-Conserved Elements (UCEs) Probe Set	Hybridization probes used to capture and sequence highly conserved genomic regions flanked by variable sequences, providing a standardized set of loci across divergent taxa [6].
RNA-Seq Library Prep Kit	For converting extracted total RNA into sequencing-ready libraries to generate transcriptome data, which is a cost-effective source for thousands of nuclear orthologous genes [7].
D-statistic Pipeline (e.g., Dsuite)	A software package specifically designed to calculate D-statistics from genome-wide variant data to test for introgression [6].
ASTRAL Software	A widely used tool for estimating the species tree from a set of input gene trees using the multi-species coalescent model, which is robust to ILS [7].
IQ-TREE Software	A software for maximum likelihood phylogenomic inference, useful for both concatenated analyses and inferring individual gene trees, and for calculating concordance factors [7].
PhyloNet	Software for modeling and analyzing phylogenetic networks, allowing for the visualization and testing of evolutionary scenarios that include reticulate events like hybridization and introgression [7].

Troubleshooting Guide: Identifying the Source of Phylogenomic Discordance

Q1: What are the primary evolutionary scenarios that lead to conflicting gene trees?

Conflicting gene trees in phylogenomic analyses predominantly arise from two biological processes: Incomplete Lineage Sorting (ILS) and Introgression. Both processes create incongruence between individual gene histories and the overall species tree, but they stem from different mechanisms and leave distinct genomic signatures.

• Incomplete Lineage Sorting (ILS): occurs when ancestral genetic polymorphisms persist through successive speciation events and are sorted inconsistently into the descendant species. This is particularly common when the time between speciation events is short relative to effective population size.
• Introgression: involves the transfer of genetic material from one species into the gene pool of another through hybridization and repeated backcrossing.

The table below summarizes the key characteristics that differentiate these processes.

Feature	Incomplete Lineage Sorting (ILS)	Introgression
Underlying Process	Stochastic sorting of ancestral polymorphisms [6]	Hybridization and backcrossing between species [11] [12]
Typical Genomic Signal	Randomly distributed discordance [6]	Localized genomic blocks [11]
Key Driver	Short internodal branches & large ancestral population size [13]	Geographic overlap and incomplete reproductive isolation [11] [6]
Phylogenetic Signal	Tree-like discordance predictable under the multispecies coalescent [14]	Reticulate network-like patterns, often between non-sister taxa [14]

Q2: During which evolutionary periods is ILS most prevalent?

ILS is not uniformly distributed across evolutionary history. It is most pronounced during specific periods, particularly rapid radiations.

• Context of Rapid Radiations: ILS is a major factor when multiple new species arise from a common ancestor over a very short evolutionary timescale. In such scenarios, the brief intervals between speciation events (short internodal branches) provide insufficient time for ancestral polymorphisms to become fixed in each new lineage [13]. This results in different loci retaining different histories of the speciation process.
• Empirical Evidence: A study on tuco-tuco rodents (Ctenomys), a genus comprising 64 species that resulted from a recent, rapid radiation about 1.3 million years ago, found that approximately 9% of loci showed signals of ILS [13]. Similarly, research on the Myotis bat genus identified rampant phylogenetic conflict, with nearly one-third of individual gene trees being discordant with the overall species tree, much of which was attributed to ILS during the initial split between Old World and New World clades [6].

Q3: When does introgression typically occur, and what facilitates it?

Introgression can occur whenever reproductively compatible species come into contact and hybridize. Its likelihood and impact are influenced by both historical and ongoing factors.

• Historical Hybridization: Many species show evidence of ancient introgression events. For example, modern humans carry introgressed DNA from archaic hominins like Neanderthals and Denisovans, a result of hybridization that occurred thousands of generations ago [11].
• Contemporary Gene Flow: Introgression is not merely a historical process. Genomic studies of New World Myotis bats have revealed signals of both historic and potential contemporary gene flow in areas where species' ranges overlap (sympatry) [6].
• Environmental Triggers: Environmental changes are a significant catalyst for introgression. Shifts in species distributions, such as those following the last glacial maximum, have created new zones of contact and opportunities for hybridization [11]. More recently, human-driven habitat changes and climate change have been observed to trigger introgression by altering species distributions and creating strong selective pressures that favor adaptive introgressed alleles [11].

Q4: What are the key methodological approaches for distinguishing ILS from introgression?

Accurately distinguishing between ILS and introgression requires a combination of phylogenetic and population genetic methods. The following workflow outlines a robust analytical strategy.

Summary of Key Methods:

• Phylogenetic Networks: Methods based on the multispecies network coalescent (MSNC), as implemented in software like PhyloNet, provide a powerful framework for inferring evolutionary histories that include both ILS and introgression simultaneously. These models do not assume a priori knowledge of the species tree [14].
• D-Statistics (ABBA-BABA Test): This is a widely used test for detecting introgression. A significant D-statistic indicates an excess of shared derived alleles between non-sister taxa, which is a classic signature of introgression [6] [13].
• Local Ancestry Inference: Methods like hidden Markov models (HMMs) and conditional random fields (CRFs) can identify specific genomic regions that have been introgressed. These tools are particularly effective for detecting recent introgression, as the introgressed segments remain long and unbroken [11].
• Retrotransposon Presence/Absence Analysis: The presence/absence patterns of SINE (Short INterspersed Element) insertions are powerful phylogenetic markers because they are homoplasy-free. Widespread discordance in these markers can be strong evidence for ILS, as they are not susceptible to the confounding effects of sequence-based selection [6].

The Scientist's Toolkit: Research Reagent Solutions

The table below lists essential bioinformatic tools and data types for investigating ILS and introgression.

Tool or Resource	Type	Primary Function
PhyloNet	Software Package	Infers phylogenetic networks from gene trees under the MSNC model, accounting for both ILS and introgression [14].
D-Statistics (ABBA/BABA)	Population Genetic Test	Detects signals of introgression by measuring allele sharing patterns between taxa [6] [13].
Hidden Markov Models (HMMs)	Statistical Model	Used for local ancestry inference to identify specific introgressed genomic regions [11].
Ves SINEs / Retrotransposons	Genomic Marker	Nearly homoplasy-free phylogenetic characters used to untangle deep evolutionary relationships and quantify ILS [6].
Transcriptomic/Exomic Data	Genomic Data	Provides sequences of thousands of orthologous genes, enabling genome-scale assessments of gene tree discordance [13].

Frequently Asked Questions (FAQs)

Q5: Can ILS and introgression occur simultaneously in the same group of organisms?

Yes, ILS and introgression are not mutually exclusive and can act simultaneously within the same clade, making phylogenetic reconstruction particularly challenging. For example, the reanalysis of the Anopheles gambiae species complex using phylogenetic networks revealed an evolutionary history shaped by multiple hybridization events (introgression) against a background of ILS [14]. Similarly, studies on Myotis bats have concluded that both ILS and gene flow have contributed significantly to the observed genomic discordance [6]. Disentangling their relative contributions requires the use of models, such as the multispecies network coalescent, that can account for both processes at once.

Q6: How does the genomic landscape of introgression differ from that of ILS?

The genomic signatures of these two processes are fundamentally different, which aids in their identification.

• Introgression leaves a localized and heterogeneous signature. Due to recombination, introgressed DNA is found in discrete blocks within the genome. The distribution of these blocks is uneven because selection acts against the introgression of alleles that are maladaptive in the recipient genetic background, creating "resistant" regions. Conversely, adaptive alleles can introgress and sweep through a population, creating peaks of high divergence [11].
• ILS produces a genome-wide and more homogeneous pattern of discordance. The incongruence between gene trees and the species tree caused by the random sorting of ancestral polymorphisms is distributed more randomly across the genome [6] [14].

Q7: What are the biggest analytical pitfalls when trying to distinguish these processes?

The most significant pitfall is misattributing the signal from one process to the other.

• Assuming a Strictly Tree-like History: Applying standard species tree inference methods that only account for ILS (like the multispecies coalescent) to data with a history of introgression can lead to inaccurate estimates of branch lengths and population sizes, as the model struggles to explain all the incongruence [14].
• Overlooking ILS in Reticulate Analyses: Conversely, parsimony-based network methods that interpret all gene tree conflict as evidence of hybridization will overestimate the number of reticulation events if ILS is a major contributor to the discordance [14].
• Reliance on Single Genomic Regions: Making conclusions based on a single marker (e.g., just the mitochondrial DNA) is highly prone to error, as its history may not reflect the species' history due to either ILS or introgression. Genome-scale data is essential for a confident diagnosis [6] [13].

Troubleshooting Guides

Guide 1: Resolving Incongruence Between Nuclear and Plastid Phylogenies

Problem: Your nuclear gene tree and plastid gene tree show strongly supported but conflicting topologies for the same taxa, making species relationships unclear.

Diagnosis: This conflict typically indicates either Incomplete Lineage Sorting (ILS) or introgression. ILS occurs when ancestral genetic polymorphisms persist through speciation events, while introgression involves transfer of genetic material between species through hybridization [7] [1].

Solution Steps:

• Calculate Site Concordance Factors - Quantify the proportion of supporting sites for each branch to assess ILS influence [7].
• Perform D-Statistics (ABBA-BABA) Tests - Test for significant allele sharing patterns indicative of introgression [15].
• Construct Phylogenetic Networks - Use methods like QuIBL to model both ILS and introgression simultaneously [7].
• Compare Multiple Gene Trees - Analyze patterns across numerous nuclear orthologous genes to distinguish random discordance (ILS) from directional discordance (introgression) [1].

Expected Outcomes:

• Balanced discordant gene trees ≈ ILS
• Significantly unbalanced discordance ≈ Introgression
• Network models with horizontal edges ≈ Confirm introgression

Guide 2: Addressing False Positive Introgression Signals from Rate Variation

Problem: D-statistics indicate introgression, but you suspect false positives due to evolutionary rate variation among lineages.

Diagnosis: Substitution rate variation across lineages can create homoplasies that mimic introgression signals in site pattern tests [15].

Solution Steps:

• Test for Rate Variation - Check for significant branch length differences across gene trees.
• Apply Clustering-Based Tests - Use methods that detect genuine introgression through spatial clustering of introgressed sites along genomes [15].
• Use Tree-Based D-Statistics - Analyze tree topologies rather than site patterns to reduce homoplasy effects [15].
• Simulate Validation - Generate expected patterns under rate variation alone to compare with observed data.

Critical Check: Genuine introgression tracts cluster genomically; homoplasy-based false positives distribute randomly [15].

Frequently Asked Questions

Q: What is the minimum sampling required to detect introgression versus ILS? A: You need genomic data from at least three ingroup species and one outgroup. For rooted triplets, this enables D-statistics and gene tree frequency analyses that can distinguish the balanced discordance of ILS from the unbalanced patterns of introgression [1].

Q: Can we accurately detect very ancient introgression events? A: Detection becomes challenging for ancient events due to recombination breakdown and potential rate variation effects. Studies have reported detectable introgression from 11-46 million years in various groups, but methodological limitations exist for older events [15]. Tree-based methods and new clustering tests improve ancient introgression detection.

Q: How do we handle non-monophyletic species in phylogenetic networks? A: Non-monophyly often indicates either ILS or recent introgression. In the Tulipa case study, most traditional sections were non-monophyletic, requiring network analyses to distinguish these causes. Implement polytomy tests and examine gene tree distributions around the problematic nodes [7].

Q: What are the limitations of D-statistics for deep divergences? A: D-statistics assume constant evolutionary rates and minimal homoplasy, which often violates in deep divergences. Rate variation creates false positives, while homoplasy can mask true signals. Supplement with tree-based methods and branch length analyses [15].

Experimental Protocols & Data Presentation

Table 1: Quantitative Comparison of Phylogenomic Conflict Detection Methods

Method	Data Requirement	Detection Power	Key Limitations	Best Application Context
D-Statistic (ABBA-BABA)	Genome-wide SNP data or sequenced loci	High for recent introgression	False positives from rate variation; requires specified species relationships [15]	Testing specific introgression hypotheses between closely-related taxa
Phylogenetic Networks	Multiple loci or genome-wide data	High for visualizing multiple processes	Computational intensity; model selection challenges [7]	Modeling complex evolutionary histories with both ILS and introgression
Site Concordance Factors	Aligned sequence data across many loci	Quantifies ILS influence	Does not directly detect introgression [7]	Assessing confidence in tree branches and ILS prevalence
QuIBL Analysis	Time-calibrated trees and genomic data	Can quantify timing of introgression	Requires careful parameterization and model testing [7]	Dating introgression events and comparing alternative histories
Tree-Based D-Statistic	Locus-specific gene trees	More robust to homoplasy than site-based tests	Dependent on accurate gene tree estimation [15]	Deep divergences where homoplasy is concerning

Table 2: Essential Research Reagents & Computational Tools

Research Reagent/Tool	Function	Application in ILS/Introgression Research
Transcriptome Sequencing	Generates nuclear orthologous genes	Provides data for constructing nuclear phylogenies and detecting discordance [7]
Whole Plastome Data	Provides uniparental inheritance signal	Serves as reference against nuclear patterns to detect cytoplasmic capture [7]
ASTRAL Software	Species tree inference under MSC	Estimates species trees accounting for ILS [7]
DSuite Package	Implements D-statistics and related tests	Detects introgression from genome-wide data [15]
ggtree R Package	Phylogenetic tree visualization and annotation	Enables effective visualization of complex phylogenetic relationships [16]
Phylo-color Script	Adds color to phylogenetic tree nodes	Facilitates visual tracking of taxa and clades in complex trees [17]

Protocol 1: D-Statistic Implementation for Introgression Detection

Purpose: Test for significant introgression between non-sister taxa using genome-wide data.

Workflow:

• Data Preparation - Generate whole genome sequences or numerous sequenced loci for four taxa: P1, P2, P3, and outgroup O.
• Variant Calling - Identify bi-allelic sites across all taxa.
• Pattern Counting - Tally "ABBA" and "BABA" sites where:
  • ABBA: P1 and outgroup share ancestral allele, P2 and P3 share derived allele
  • BABA: P1 and P3 share derived allele, P2 and outgroup share ancestral allele [15]
• Statistical Testing - Calculate D = (ABBA - BABA) / (ABBA + BABA) and assess significance with block jackknife.

Interpretation: Significant D ≠ 0 indicates excess allele sharing between P3 and either P1 or P2, suggesting introgression [15] [1].

Protocol 2: Phylogenomic Network Construction

Purpose: Reconstruct evolutionary histories involving both vertical descent and horizontal introgression.

Workflow:

• Gene Tree Estimation - Infer trees for hundreds to thousands of nuclear orthologous genes [7].
• Discordance Analysis - Calculate site concordance factors (sCF) and discordance factors (sDF1/sDF2) to identify nodes with high or imbalanced discordance [7].
• Network Inference - Apply network methods to nodes showing significant discordance not explained by ILS alone.
• Model Testing - Compare network models with different introgression scenarios using statistical criteria.

Visualization Workflows

Diagram 1: Phylogenomic Conflict Resolution Workflow

Diagram 2: D-Statistic Signal Detection Logic

Implications for Species Delimitation and Evolutionary History Reconstruction

Technical Support Center: Troubleshooting Phylogenomic Analysis

Frequently Asked Questions (FAQs)

FAQ 1: My phylogenetic analyses show widespread gene tree discordance. How can I determine if it is caused by Incomplete Lineage Sorting (ILS) or introgression?

Answer: Widespread gene tree discordance can indeed be caused by both ILS and introgression. To distinguish between them, you should:

• Calculate Gene Tree Frequencies: Under a pure ILS scenario, the two discordant gene tree topologies are expected to be equal in frequency for a rooted triplet of species [1]. A significant deviation from this 1:1 ratio, where one discordant topology is more frequent, is a key statistical signature of introgression [1].
• Use Summary Statistics: Apply the D-statistic (ABBA-BABA test) to biallelic site patterns. A significant D-statistic signal indicates an excess of shared derived alleles between two species, which is inconsistent with ILS alone and suggests introgression [1].
• Analyse Branch Lengths: Compare branch lengths on gene trees. Introgression can lead to shorter branches between the introgressing lineages, a pattern not typically expected under ILS [1].

FAQ 2: For a reliable D-statistic test, what is the minimum sampling requirement and what are common pitfalls?

Answer:

• Minimum Sampling: The minimal requirement for a powerful D-statistic test is genomic data from a single haploid individual from each of three focal species (P1, P2, P3) and an outgroup (O), forming an unrooted quartet [1].
• Common Pitfalls:
  • Incorrect Outgroup: Using an outgroup that is too distant or too closely related can lead to incorrect polarization of ancestral and derived alleles.
  • Multiple Introgression Events: Complex histories with multiple pulses of gene flow can cancel out signals and lead to false negatives.
  • Gene Flow from an Unsampled Lineage: If the source of introgression is not included in your analysis ("ghost" introgression), results can be misinterpreted.

FAQ 3: My data suggests introgression is present. How can I characterize the direction, timing, and extent of the introgression event?

Answer: Characterizing introgression requires moving beyond simple detection.

• Direction: The direction of gene flow (e.g., P3 → P2) can often be inferred from the asymmetrical frequencies of gene tree topologies or from methods like the D-statistic, which tests for an excess of allele sharing between a specific pair of species [1].
• Extent: The fraction of the genome affected by introgression can be estimated by calculating the proportion of loci that show a strong signal of introgression or by using methods like fd to quantify the local ancestry of genomic blocks [18].
• Timing: More advanced, model-based approaches such as phylogenetic network inference (e.g., using tools like PhyloNet or SNIPPY) can co-estimate the species tree and introgression events, providing estimates of the timing and magnitude of gene flow [1].

FAQ 4: What are the limitations of using a single sample per species in phylogenomic studies of introgression?

Answer: While many phylogenomic methods are designed for one sample per species and are robust for detection [1], this approach has limitations for characterization:

• Underestimation of Diversity: A single sample may not capture the full genetic diversity within a species, which can bias estimates of population genetic parameters.
• Inability to Detect Recent Gene Flow: Recent introgression that has not become fixed in the population might be missed if the introgressed haplotype is not present in the single sampled individual.
• Ongoing Speciation: High introgression levels between defined "species" may indicate that they are not yet fully isolated and are in the process of speciation. Using multiple samples can help test this by revealing if gene flow is widespread or restricted to specific sub-populations [18].

Experimental Protocols for Key Analyses

Protocol 1: D-Statistic (ABBA-BABA) Test

1. Objective: To test for asymmetry in allele sharing patterns that indicates introgression between two sister species (P2 and P3) against an ILS null hypothesis.

2. Methodology:

• Data Preparation: Generate a whole-genome multiple sequence alignment for four taxa: P1, P2, P3, and an outgroup (O).
• Variant Calling: Identify biallelic sites across the genome. For each site, polarize alleles as ancestral (A) or derived (B) using the outgroup.
• Site Pattern Counting: Count the occurrences of two specific site patterns across the genome:
  • ABBA: Sites where P2 and P3 share a derived allele not found in P1.
  • BABA: Sites where P1 and P3 share a derived allele not found in P2.
• Calculation: Compute the D-statistic using the formula:
  • D = (NABBA - NBABA) / (NABBA + NBABA)
• Significance Testing: Assess the statistical significance of the D-value using a block jackknife or binomial test. A significant deviation from zero indicates introgression.

3. Interpretation of Results:

• D > 0: Suggests introgression between P3 and P2.
• D < 0: Suggests introgression between P3 and P1.
• D ≈ 0: Consistent with the null hypothesis of no introgression (discordance is likely due to ILS).

Protocol 2: Phylogenetic Network Inference with SNIPPY

1. Objective: To infer a phylogenetic network that explicitly models introgression events, estimating their direction and weight.

2. Methodology:

• Input Data: A concatenated alignment of multiple loci or a set of gene trees from the species of interest.
• Model Selection: Specify a set of candidate species networks that may explain the data.
• Likelihood Calculation: For each candidate network, the tool calculates the likelihood of the observed gene trees or sequence data under the multispecies network coalescent model.
• Parameter Estimation: The method co-estimates the species phylogeny and introgression events, providing parameters like the probability of gene flow (introgression weight) and the direction of flow.

3. Interpretation of Results:

• The inferred network will display a reticulation node, which represents the hybridization/introgression event.
• The introgression weight (γ) on the reticulation edge indicates the proportion of the genome that originated from the donor lineage, helping to quantify the extent of introgression [1].

Table 1: Prevalence of Introgression in Bacterial Core Genomes Across Select Genera This table summarizes quantitative findings on introgression levels, illustrating the variability of this process across different lineages [18].

Bacterial Genus / Lineage	Average % of Introgressed Core Genes	Maximum % Observed (and Species)	Key Contextual Factor
Escherichia–Shigella	Information Not Specified	~14%	Highest level among the 50 lineages studied [18].
Cronobacter	Information Not Specified	High (specific % not stated)	Listed among the genera with the highest levels [18].
Across 50 Genera (Average)	~8% (Mean), ~3% (Median)	-	Introgression is common but highly variable [18].
Streptococcus parasanguinis	33.2% (between ANI-sp32 & ANI-sp67)	-	Later classified as a single species, showing how definition impacts estimates [18].

Table 2: Key Research Reagent Solutions for Phylogenomic Analysis This table lists essential software tools and data types used for detecting and characterizing introgression.

Item Name	Type	Primary Function in Analysis
D-Statistic	Software Script / Method	A hypothesis-testing method to detect introgression by testing for asymmetry in allele sharing patterns [1].
PhyloNet / SNIPPY	Software Package	Model-based programs for inferring phylogenetic networks and explicitly estimating introgression parameters from multi-locus data [1].
Whole-Genome Sequencing Data	Data Type	Provides the high-density genomic markers (SNPs, full sequences) needed to infer gene trees and detect introgressed loci [1] [18].
Multi-Locus Sequence Alignment	Data Type	A formatted dataset of aligned DNA sequences from multiple loci across multiple individuals, the fundamental input for phylogenetic tree and network estimation [18].

Workflow Visualization

Analytical Toolkit: Statistical Methods and Computational Approaches for Signal Discrimination

What is the ABBA-BABA Test?

The ABBA-BABA test, also known as Patterson's D-statistic, provides a powerful method for detecting deviations from a strictly bifurcating evolutionary history, most commonly used to test for introgression using genome-scale SNP data [19]. This method compares the frequencies of two discordant site patterns ("ABBA" and "BABA") that arise when gene genealogies differ from the species tree due to processes like introgression or incomplete lineage sorting (ILS) [20].

The test operates on a four-taxon system with an established phylogeny: (((P1, P2), P3), O), where P1, P2, and P3 are ingroup populations and O is an outgroup. The core principle is that under a strict bifurcating tree with no gene flow, the two discordant genealogical patterns ABBA and BABA should occur with roughly equal frequency. A significant deviation from this 1:1 ratio indicates potential introgression [19] [20].

Scientific Foundations and Interpretation

D-Statistic Calculation: The D-statistic is calculated as: D = (ΣABBA - ΣBABA) / (ΣABBA + ΣBABA) [19]

Where:

• ABBA sites are those where P2 and P3 share a derived allele ('B') while P1 has the ancestral state ('A')
• BABA sites are those where P1 and P3 share the derived allele while P2 has the ancestral state [19]

Interpretation Guidelines:

• D = 0: No significant deviation, consistent with no gene flow
• D significantly positive: Excess of ABBA sites, suggesting introgression between P2 and P3
• D significantly negative: Excess of BABA sites, suggesting introgression between P1 and P3 [20]

Statistical significance is typically assessed using a Z-score, where |Z| > 3 is considered significant, corresponding to a p-value of approximately 0.001 [20].

Table 1: D-Statistic Interpretation Guide

D Value	Z-Score	Interpretation	Suggested Conclusion
≈ 0		Z	< 3	No significant deviation	No evidence of gene flow
Significantly > 0	Z ≥ 3	Excess ABBA sites	Possible gene flow between P2 and P3
Significantly < 0	Z ≤ -3	Excess BABA sites	Possible gene flow between P1 and P3

Experimental Design and Workflow

Standard Analysis Workflow

The following diagram illustrates the comprehensive workflow for conducting ABBA-BABA analysis, from data preparation through interpretation:

Essential Research Reagents and Tools

Table 2: Essential Software Tools for ABBA-BABA Analysis

Tool Name	Primary Function	Input Format	Key Features	Citation
Dsuite	Fast D-statistics and f4-ratio	VCF	Efficient genome-scale calculations across all population combinations	[21] [22]
ipyrad	Population genomics analysis	Loci data	Tree-based hypothesis testing with visualization	[23]
ANGSD	ABBABABA analysis	BAM	Works with low-depth NGS data, no called genotypes required	[24]
R/Python Scripts	Custom analysis	Frequency tables	Flexible for specific research needs	[19]

Table 3: Required Input Files and Specifications

File Type	Format	Essential Content	Purpose
Genotype Data	VCF, BAM, or genotype tables	Bi-allelic SNPs for all individuals	Primary genetic data input
Population Map	Text file (tab-delimited)	Individual → Population assignments	Define populations for analysis
Tree File	Newick format (optional)	Phylogenetic relationships	Guide hypothesis testing
Outgroup Sequence	FASTA or specified in VCF	Ancallelle information	Polarize alleles as ancestral/derived

Troubleshooting Common Experimental Issues

Data Quality and Preparation Problems

Problem: Inconsistent results when using different software tools

• Potential Cause: Different default parameters, filtering criteria, or statistical implementations
• Solution:
  • Standardize input data quality filters (minimum depth, missing data thresholds)
  • Compare results across multiple tools (Dsuite, ipyrad, ANGSD)
  • Validate with known positive controls from published datasets [22] [24]

Problem: Low number of informative sites (ABBA + BABA)

• Potential Cause: Insufficient genomic coverage, poor outgroup choice, or excessive filtering
• Solution:
  • Increase genomic sampling (more loci or whole genomes)
  • Verify outgroup appropriateness (should be fixed for ancestral allele)
  • Relax filtering thresholds while maintaining quality [19] [25]

Problem: Missing data causing biased estimates

• Potential Cause: Uneven sequencing coverage across samples
• Solution:
  • Implement minimum coverage thresholds (e.g., DP ≥ 10 in ANGSD)
  • Use methods that account for uncertainty (ANGSD with genotype likelihoods)
  • Apply population-level allele frequency estimation rather than requiring fixed differences [24]

Analysis and Interpretation Challenges

Problem: Significant D-statistic but uncertain if due to introgression or ILS

• Potential Cause: Both processes can generate similar patterns of genealogical discordance
• Solution:
  • Perform multiple tests across different population combinations
  • Use additional statistics like f4-ratio to quantify admixture proportions
  • Conduct phylogenetic network analyses to evaluate alternative explanations [26] [7]
  • Apply Dsuite's Fbranch method to interpret systems of f4-ratio results [22]

Problem: Weak statistical support (low Z-scores) despite large dataset

• Potential Cause: Inadequate block jackknife implementation or strong linkage disequilibrium
• Solution:
  • Adjust block size (typically 1-5 Mb for humans, larger for low-recombination regions)
  • Ensure sufficient number of jackknife blocks (≥20 recommended)
  • Verify block independence exceeds LD decay distance [19] [24]

Problem: Direction of introgression unclear

• Potential Cause: D-statistic indicates gene flow but doesn't specify direction
• Solution:
  • Implement Dsuite's Dinvestigate for fd, fdM, and df statistics
  • Use Five-taxon Dfoil tests for more complex scenarios
  • Consider additional methods like QuIBL for directionality assessment [26] [7]

Advanced Applications in Distinguishing ILS from Introgression

Multi-Method Approaches for Complex Scenarios

The following diagram illustrates the logical decision process for distinguishing incomplete lineage sorting (ILS) from introgression:

Case Studies: Successful ILS-Introgression Distinction

Case Study 1: Tuco-tucos (Ctenomys) Radiation

• Challenge: Rapid radiation with both ILS and introgression signals
• Approach: Combined D-statistics with Dfoil analysis
• Finding: Approximately 9% of loci showed ILS patterns, with significant introgression detected from C. torquatus into C. brasiliensis
• Methodological Insight: Five-taxon tests (Dfoil) provided additional resolution beyond standard four-taxon D-statistics [26]

Case Study 2: Liliaceae Tribe Tulipeae

• Challenge: Pervasive phylogenetic incongruence among Tulipa, Amana, and Erythronium
• Approach: Transcriptome sequencing with D-statistics and QuIBL analysis
• Finding: Both ILS and reticulate evolution contributed to discordance, with D-statistics helping identify specific introgression events
• Methodological Insight: Site concordance factors (sCF) and discordance factors (sDF) helped quantify conflicting signals [7]

Case Study 3: Gossypium (Cotton) Adaptive Radiation

• Challenge: Complex speciation history with rapid diversification
• Approach: Whole-genome analysis with detailed ILS mapping
• Finding: Non-random distribution of ILS regions with evidence of natural selection
• Methodological Insight: Integration of D-statistics with selection scans revealed adaptive significance of some ILS regions [27]

Frequently Asked Questions (FAQs)

Methodological Questions

Q: How many samples are needed per population for reliable D-statistic analysis? A: While single samples can be used, multiple individuals per population are recommended for robust allele frequency estimation. The method can incorporate frequency information from multiple individuals, increasing power and reliability [19] [22].

Q: What genetic distance is appropriate for ABBA-BABA tests? A: The D-statistic is robust across a wide range of genetic distances but is most effective for closely to moderately diverged taxa. Studies have successfully applied it to taxa with sequence divergences from 0.3% to 4-5% [25].

Q: How should block size be determined for jackknife resampling? A: Block size should exceed the distance at which linkage disequilibrium decays to background levels. For humans, 5 Mb is commonly used. For other organisms, estimate LD decay from your data or use conservative larger blocks [19] [24].

Interpretation Questions

Q: Can a significant D-statistic alone prove introgression? A: No. While a significant D-statistic indicates genealogical discordance, other processes including ancestral population structure, selection, or among-species rate variation can also produce significant results. Always consider alternative explanations and use complementary methods [20] [25].

Q: How can I distinguish recent from ancient introgression? A: Recent introgression typically shows stronger clustering of ABBA-BABA signals along chromosomes, while ancient introgression is more dispersed. Dsuite's --ABBAclustering option specifically tests for such clustering patterns [22].

Q: What proportion of the genome needs to be introgressed for detection? A: The detectable proportion depends on population sizes, divergence times, and number of sites analyzed. Simulation studies suggest the method can detect introgression affecting as little as 1-5% of the genome with sufficient data [25].

Technical Implementation Questions

Q: How do I handle missing data in ABBA-BABA analyses? A: Most modern implementations (Dsuite, ANGSD) can handle missing data by estimating allele frequencies from available data. Avoid excessive missingness, and consider using methods that incorporate genotype uncertainty rather than simple missingness thresholds [24] [22].

Q: What are the computational requirements for genome-scale D-statistic analysis? A: Requirements vary by tool. Dsuite is optimized for efficiency with large VCF files. For 100 whole genomes, analyses typically require moderate computational resources (8-16 GB RAM, hours to days runtime). Memory scales with number of populations and SNPs [21] [22].

Q: Can I perform ABBA-BABA tests without an outgroup? A: Traditional D-statistics require an outgroup. However, the recently developed D3 statistic uses genetic distances instead of ancestral allele identification, circumventing the need for an outgroup [20]. Dsuite also offers Dquartets for quartet-based analysis without an outgroup [22].

Troubleshooting Guides

Diagnosing Phylogenetic Discordance

Table 1: Diagnosing Sources of Gene Tree Discordance

Observed Pattern	Potential Cause	Diagnostic Tests	Recommended Solutions
Widespread, random discordance among gene trees, especially near short internal branches.	Incomplete Lineage Sorting (ILS)	- Calculate site concordance factors (sCF) [7]- Perform polytomy tests [7]- Use quartet-based measures like sCF and sDF1/sDF2 [7]	- Apply Multi-Species Coalescent (MSC) models (e.g., ASTRAL) [7]- Increase the number of independent loci [6]
Discordance concentrated in specific genomic regions or taxa; evidence of allele sharing between non-sister taxa.	Introgression (Hybridization)	- D-statistics (ABBA-BABA tests) [7] [6]- Phylogenetic network analysis [28] [29]- Compare parapatric vs. allopatric populations [30]	- Construct phylogenetic networks (e.g., using Quartet-based methods) [28] [29]- Use explicit network inference tools (e.g., QuIBL) [7]
Strong conflict between nuclear and organelle (e.g., plastid) phylogenies.	Reticulate Evolution (e.g., Hybridization) or Past Gene Flow	- Concordance factor analysis [7]- D-statistics on different genomic compartments [7]	- Analyze nuclear and plastid datasets separately to identify conflicting signals [7]- Use models that account for different inheritance patterns [30]
Multiple, equally optimal trees with no clear dominant signal.	Simultaneous Divergence (Hard Polytomy) or Data Insufficiency	- Polytomy tests [7]- Evaluate statistical support for bipartitions [31]	- Increase genomic sampling [7]- Use methods designed for rapid radiations (e.g., SINEs) [6]
Gene tree discordance that is not random and correlates with specific traits or geography.	Gene Flow	- Population structure analysis (e.g., using programs like STRUCTURE or ADMIXTURE) [30]- Approximate Bayesian Computation (ABC) [30]	- Model demographic history with ABC [30]- Use ecological niche modeling to test for secondary contact [30]

Resolving Technical and Analytical Challenges

Table 2: Solving Common Network Inference Problems

Problem	Root Cause	Solutions & Best Practices
Network is overly complex with too many reticulations.	Over-fitting to noise or ILS rather than true introgression.	- Use statistical tests (e.g., D-statistics) to confirm introgression before modeling it [6]- Apply methods that distinguish level-1 from level-2 networks [28] [29]- Use model selection criteria to choose the simplest adequate network.
Software fails to converge or produces errors on large datasets.	Computational limitations or model violations.	- Reduce dataset complexity by filtering orthologous genes [7]- Ensure data meets model assumptions (e.g., no recombination within loci)- Use quartet-based summary methods (e.g., from concordance factors) which are less computationally intensive [28].
Inconsistent results from different analysis methods (e.g., ML vs. Bayesian).	Different methods have varying sensitivities to ILS, gene flow, and model misspecification.	- Compare results from multiple methods (e.g., ML, MSC, network analyses) to identify robust patterns [7]- Use coalescent-based species tree methods to account for ILS [7].
Low statistical support for key nodes or reticulations.	Insufficient phylogenetic signal or high levels of conflict.	- Increase the number of informative loci (e.g., use thousands of nuclear orthologous genes) [7]- Calculate metrics like sCF to assess the per-site support for a node [7].

Frequently Asked Questions (FAQs)

Q1: What is the fundamental difference between Incomplete Lineage Sorting (ILS) and introgression?

Both processes create incongruence between gene trees and the species tree, but their mechanisms differ. ILS is a passive process resulting from the retention and random sorting of ancestral genetic polymorphisms across successive speciation events. This is particularly common in rapid radiations where short time intervals between speciations prevent alleles from reaching fixation [6] [30]. Introgression, conversely, is an active process involving the transfer of genetic material from one species into the gene pool of another through hybridization and backcrossing [30]. While ILS produces a largely random distribution of discordant gene trees, introgression generates a directional and often localized signal of allele sharing between specific taxa [6].

Q2: How can I practically distinguish whether ILS or introgression is causing gene tree discordance in my dataset?

A multi-pronged approach is necessary:

• D-statistics (ABBA-BABA tests): This is a primary test for introgression. A significant D-statistic indicates an excess of shared derived alleles between non-sister taxa, which is a hallmark of introgression [7] [6].
• Site Concordance Factors (sCF): If the supporting signal for a branch is spread randomly across many loci and sites (low sCF), it is consistent with ILS. If the signal is concentrated or there is an imbalance in discordance factors (sDF), it can suggest introgression [7].
• Geographic Sampling: Compare allopatric (geographically separated) and parapatric (adjacent) populations of the studied species. A stronger signal of allele sharing in parapatry points towards recent or ongoing gene flow, while a uniform signal across geographic samples is more indicative of ILS [30].
• Phylogenetic Networks: Use network models that can visually represent conflicting signals as reticulations. Tools that infer networks from quartet concordance factors are theoretically grounded for this purpose [28] [29].

Q3: My plastid (or mitochondrial) DNA tree strongly conflicts with my nuclear species tree. Which one should I trust?

This is a classic signature of reticulate evolution. You should not inherently "trust" one over the other; instead, you should investigate the cause of the conflict. Organelle genomes are often maternally inherited and can have different evolutionary histories than the nuclear genome due to past hybridization events (chloroplast capture) [7] [30]. The nuclear genome, being biparentially inherited, may represent the primary species history, while the organelle genome might reflect a history of hybridization. Analyzing both genomes in conjunction allows you to test these hypotheses.

Q4: What are the advantages of using SINE insertions or other retrotransposons for phylogenomics?

SINEs (Short INterspersed Elements) are considered nearly ideal phylogenetic markers for several reasons [6]:

• Low Homoplasy: The probability of independent, parallel insertions of a SINE at the same genomic location in different lineages is exceedingly low. They are also rarely precisely excised.
• Polarity: The absence of an insertion is the ancestral state, and its presence is the derived state. This unambiguous polarity eliminates the need for root inference.
• Insensitivity to Evolutionary Models: As presence/absence characters, they are not subject to complications of sequence evolution models, such as multiple hits and rate variation.
• Resolving Power: They are particularly powerful for resolving difficult branches in rapid radiations where ILS is high, as they are less susceptible to its effects compared to sequence SNPs [6].

Q5: My phylogenetic analysis resulted in a polytomy. Does this mean I have a "true" hard polytomy, or is it a limitation of my data?

A polytomy in a phylogenetic tree can represent either a soft polytomy, which is an unresolved node due to insufficient data or high levels of conflict (like ILS), or a hard polytomy, which implies a true simultaneous divergence of multiple lineages [7]. To distinguish between them, you can:

• Increase data: Significantly increase the number of independent loci. If the polytomy persists with massive genomic data, it becomes more plausible.
• Conduct polytomy tests: Use statistical tests to evaluate whether a resolved tree around the polytomous node fits the data significantly better than a multifurcating tree [7].
• Look for consistent short internodes: If multiple methods consistently recover very short branch lengths around the node, it supports a brief, explosive diversification.

Experimental Protocols & Workflows

A Workflow for Discriminating ILS from Introgression

The following diagram outlines a logical workflow for analyzing phylogenomic data where ILS and introgression are suspected.

Protocol: D-Statistics Analysis for Introgression Detection

Purpose: To test for evidence of gene flow between a focal species and a closely related outgroup using genomic data.

Principle: The D-statistic (or ABBA-BABA test) compares patterns of shared derived alleles between four taxa (((P1, P2), P3), Outgroup). An excess of ABBA or BABA patterns over the null expectation indicates introgression between P3 and P2 or P1, respectively [6].

Materials:

• Genomic sequence data (e.g., whole-genome resequencing, UCEs, transcriptomes) for at least four individuals/taxa in the required configuration.
• A high-quality reference genome (for alignment) or a set of orthologous sequences.
• Computational tools like ANGSD, Dsuite, or dedicated packages in R/Python.

Procedure:

• Taxon Selection: Define your four taxa carefully:
  • P1 & P2: Sister species (or populations).
  • P3: The putative introgressing taxon (tested for gene flow with P1 or P2).
  • Outgroup (O): A more distantly related taxon to polarize the alleles.

• Variant Calling: Map reads to a reference genome or align orthologous sequences. Call SNPs rigorously, filtering for quality, depth, and missing data.

• Genotype Likelihood/Count: For each SNP site, count or estimate the probabilities for the four possible allele patterns relative to the outgroup:

  • ABBA: Derived in P2 and P3, ancestral in P1.
  • BABA: Derived in P1 and P3, ancestral in P2.

• Calculate D-Statistic:

  • Use the formula: D = (NABBA - NBABA) / (NABBA + NBABA)
  • Where NABBA and NBABA are the counts of the respective sites.

• Significance Testing:

  • Perform a block jackknife or bootstrap resampling to estimate the standard error and calculate a Z-score. A |Z| > 3 is generally considered significant evidence of introgression.

• Interpretation:

  • A significantly positive D suggests gene flow between P3 and P2.
  • A significantly negative D suggests gene flow between P3 and P1.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Materials and Tools for Phylogenomic Network Analysis

Item/Reagent	Function/Application	Example Use Case
Transcriptome Sequencing (RNA-Seq)	Provides thousands of low-copy nuclear orthologous genes for phylogenomic analysis without needing a whole genome [7].	Inferring species trees and quantifying gene tree discordance in non-model organisms with large genomes (e.g., Tulipa) [7].
Ultraconserved Elements (UCEs)	Targeted sequencing of highly conserved genomic regions that flank variable sequences, useful across deep and shallow evolutionary timescales.	Phylogenetic studies of diverse groups like bats (Myotis); can be compared with retrotransposon-based phylogenies [6].
SINE (Retrotransposon) Presence/Absence Profiling	A powerful phylogenetic marker with minimal homoplasy, ideal for resolving rapid radiations and detecting deep introgression [6].	Untangling the evolutionary history of mammalian clades or bat genera with extensive ILS and hybridization [6].
Quartet Concordance Factors (CFs)	Metrics that quantify the proportion of genes supporting each of the three possible quartet topologies for a set of four taxa.	Diagnosing sources of discordance (ILS vs. introgression) and providing the input data for robust phylogenetic network inference [28] [29].
Phylogenetic Network Software (e.g., ASTRAL, methods for level-2 networks)	Software packages that implement the Multi-Species Coalescent and network models to infer species trees/networks from discordant gene trees.	Reconstructing evolutionary histories that include hybridization events, moving beyond bifurcating trees [7] [28].
Approximate Bayesian Computation (ABC)	A framework for comparing complex demographic scenarios (e.g., isolation-with-migration vs. strict isolation) to infer historical population processes.	Testing between ILS and secondary contact as explanations for shared genetic variation between pine species (Pinus) [30].

Troubleshooting Guide: Common MSC Analysis Issues

Problem 1: Unexpected Parameter Estimates in Species Tree Inference

Symptoms: Inferred population sizes are consistently underestimated or divergence times are inflated compared to known values.

Diagnosis: This pattern often indicates a violation of model assumptions. A recent 2023 study demonstrated that intra-locus recombination, even at realistic biological levels, can cause these specific estimation errors. When recombination breaks sequences into smaller effective coalescent units, methods assuming non-recombining loci can produce biased parameter estimates [32].

Solution:

• Verify data partitioning: Ensure your loci represent genuine "coalescent genes" (c-genes) rather than just "molecular genes" (m-genes). C-genes are segments between recombination events that share the same phylogenetic history, while m-genes may contain multiple recombination breakpoints [32].
• Consider alternative methods: For genome-scale data, methods like SNAPP that use unlinked biallelic markers (SNPs) are generally robust to recombination effects since recombination cannot occur within a single site [32].
• Assess recombination rate: Use recombination detection tools (e.g., from msprime) to characterize the mean length of c-genes in your dataset, as shorter c-genes indicate higher recombination rates that may impact analysis [32].

Problem 2: False Positive Introgression Signals in Shallow Phylogenies

Symptoms: Significant D-statistics or HyDe results suggesting gene flow, but without biological evidence for hybridization.

Diagnosis: Lineage-specific rate variation can create ABBA-BABA asymmetry that mimics introgression signals. A 2025 study demonstrated that even minor rate variations (17-33% difference between sister lineages) in shallow phylogenies can inflate false positive rates up to 35-100% with 500 Mb of data [33].

Solution:

• Perform relative rate tests: Quantify rate differences between sister lineages before introgression testing [33].
• Use branch-length informed methods: Supplement site-pattern methods (D-statistic, HyDe) with methods that utilize gene-tree branch lengths (D3, QuIBL) or full-likelihood approaches that are less susceptible to rate variation artifacts [33].
• Validate with independent evidence: Corroborate significant results with demographic modeling or geographic evidence of species contact zones.

Problem 3: Poor Performance with Genomic-Scale Data

Symptoms: Method fails to converge or produces anomalous results with whole-genome sequences.

Diagnosis: Different MSC implementations have varying scalability and robustness to recombination. Surprisingly, methods specifically designed for recombination like diCal2 may perform worse than other approaches due to extensive algorithmic approximations [32].

Solution:

• Method selection: StarBEAST2 using short or medium-sized loci has shown robustness to realistic recombination rates [32].
• Data subsampling: For very large datasets, use careful subsampling strategies and verify consistency across subsets.
• Software validation: Check that your software version includes the latest bug fixes for genomic-scale analyses.

Frequently Asked Questions (FAQs)

Q: When should I use MSC methods instead of concatenation? A: MSC methods are particularly important when analyzing closely related species with short internal branches, where incomplete lineage sorting (ILS) is pervasive. Simulation studies reveal that concatenation can produce spuriously confident yet conflicting results in regions of parameter space where MSC models perform well, especially when subjected to data subsampling [34].

Q: How does the multispecies coalescent with recombination (MSC-R) differ from standard MSC? A: The MSC-R extends the MSC to explicitly include recombination processes, integrating over gene histories and recombination breakpoints. However, current implementations like diCal2 may introduce approximations that impact parameter estimation accuracy compared to methods that assume recombination-free loci [32].

Q: Can I use transcriptomic data for MSC analysis? A: Yes, but with important caveats. Exons in transcriptomes often span large chromosomal regions with substantial recombination, potentially violating the assumption of non-recombining loci. Careful partitioning of coding sequences into smaller coalescent units may be necessary for accurate inference [34].

Q: What are the key assumptions of MSC models that are most commonly violated? A: The most problematic assumptions include: (1) no recombination within loci, (2) neutral evolution, (3) correct gene tree rooting, and (4) accurate sequence alignment. Violations of these assumptions can lead to biased parameter estimates, though some methods show robustness to moderate violations [34].

Q: How can I distinguish ILS from introgression in practice? A: Use multiple complementary approaches:

• Compare D-statistics across different taxon combinations
• Examine phylogenetic network patterns for reticulate evolution
• Assess whether discordance patterns are consistent across the genome
• Use full-likelihood methods that jointly model both processes
• Remember that rate variation alone can create false introgression signals [33]

Table 1: Method Performance Under Model Violations

Method	Data Type	Robust to Recombination?	Robust to Rate Variation?	Best Use Case
StarBEAST2	Locus sequences	Yes (short/medium loci) [32]	Limited data	Divergence time estimation with moderate ILS
SNAPP	Unlinked SNPs	Yes (inherently) [32]	Moderate	Species tree topology with high ILS
diCal2	Whole genome	No (performs worse) [32]	Not assessed	Not recommended based on current evidence [32]
D-statistic	Site patterns	Varies	No (high false positives) [33]	Initial screening with rate validation

Table 2: Impact of Rate Variation on Introgression Detection (Shallow Phylogenies)

Rate Variation	Tree Depth (generations)	False Positive Rate	Recommended Mitigation
Weak (17% difference)	3×10⁵	Up to 35% [33]	Use branch-length methods
Moderate (33% difference)	3×10⁵	Up to 100% [33]	Validate with multiple methods
Any detectable	<10⁶	Significant inflation [33]	Perform relative rate test first

Experimental Protocols

Protocol 1: Simulation Framework for Testing MSC Method Robustness

Based on the whole-genome simulation approach used in recent method comparisons [32]:

Materials:

• msprime version 1.0.2 or later for coalescent simulations
• Genomic sequence data or parameters reflecting your study system
• Computing cluster access for large-scale analyses

Methodology:

• Parameterize simulation framework: Set population sizes, divergence times, mutation rates, and recombination rates reflecting your biological system. For mammalian-scale analyses, use Ne ≈ 50,000-100,000 and recombination rates of 1×10⁻⁸ to 1×10⁻⁹ events per base per generation [32].

• Generate sequence data: Use msprime to simulate whole genomes under the coalescent with recombination model. The command structure typically includes:

• Partition data: Split simulated genomes into loci of varying lengths (500bp-10kb) to test sensitivity to locus length assumptions.

• Run inference methods: Analyze the same simulated datasets with multiple MSC methods (e.g., StarBEAST2, SNAPP) and concatenation for comparison.

• Assess performance: Compare estimated parameters (divergence times, population sizes) to known simulation values using mean squared error and bias metrics.

Protocol 2: Validating Introgression Signals Against Rate Variation Artifacts

Adapted from rigorous testing protocols for distinguishing true introgression from rate variation artifacts [33]:

Materials:

• Multi-sequence alignment for at least 4 taxa (((P1,P2),P3),O)
• Computing environment with D-statistic and HyDe implementation
• Relative rate testing software

Methodology:

• Perform relative rate tests: Quantify rate variation between sister lineages P1 and P2 using outgroup O. Significant rate differences (≥10-15%) warrant caution in interpreting introgression signals [33].

• Calculate D-statistics: Compute ABBA and BABA site pattern counts and D-values using established packages. Use block jackknifing for significance testing.

• Conduct HyDe analysis: Run HyDe to test for hybrid speciation scenarios, specifying the appropriate outgroup.

• Supplement with branch-length methods: Apply methods like D3 or QuIBL that utilize branch length information and are less susceptible to rate variation artifacts.

• Interpret holistically: Only conclude introgression when multiple methods converge on significant results and rate variation has been accounted for. For shallow phylogenies (<1 million generations), be particularly cautious of false positives [33].

Method Selection Workflow

MSC Method Selection Workflow

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Computational Tools for MSC Analysis

Tool/Resource	Function	Application Context	Key Considerations
msprime	Coalescent simulation with recombination	Generating synthetic data for method validation	Essential for testing method robustness to model violations [32]
StarBEAST2	Bayesian species tree inference	Estimating divergence times and population sizes	Robust to realistic recombination rates with appropriate locus length [32]
SNAPP	Species tree from SNP data	Topology inference with high ILS	Unaffected by recombination; uses biallelic markers [32]
D-statistic implementation (e.g., Dsuite)	Introgression detection	Initial screening for gene flow	Validate against rate variation artifacts [33]
Relative rate test packages	Quantifying lineage-specific rate variation	Quality control before introgression testing	Critical for shallow phylogenies to prevent false positives [33]
Phylogenetic network software (e.g., PhyloNet)	Modeling reticulate evolution	Distinguishing ILS from introgression	Provides visual representation of conflicting signals

Frequently Asked Questions (FAQs)

FAQ 1: What are the primary sources of gene tree discordance I might encounter in my phylogenomic analysis?

Gene tree discordance, where gene trees differ from each other and from the species tree, arises from multiple sources. These can be broadly categorized into biological processes and analytical artifacts.

• Biological Sources: Incomplete Lineage Sorting (ILS) is a primary source, occurring when ancestral genetic polymorphisms fail to coalesce (merge) in the immediate ancestor of two species, causing gene trees to reflect a history deeper than the species split. Hybridization/Introgression is another key source, where gene flow between species leads to some genes having a history that differs from the species tree. Other biological causes include gene duplication and loss and horizontal gene transfer [5] [35] [36].
• Analytical Sources: These include gene-tree-inference error due to insufficient phylogenetic signal or model misspecification, hidden paralogy (when paralogous genes are mistakenly treated as orthologs), and issues like alignment errors, substitutional saturation, and compositional heterogeneity [5] [37] [35]. Distinguishing between these sources, particularly ILS and introgression, is a central goal of phylogenomic conflict analysis.

FAQ 2: How can I distinguish between incomplete lineage sorting (ILS) and introgression as causes of conflict?

Disentangling ILS from introgression requires a combination of tests and approaches, as no single method is foolproof. The table below summarizes key strategies:

Table 1: Strategies for Distinguishing ILS from Introgression

Strategy	Description	Expected Pattern for ILS	Expected Pattern for Introgression
Site Pattern Tests	Uses statistical tests like D-statistics (ABBA-BABA) to assess asymmetry in allele sharing between taxa [5].	No significant excess of allele sharing between non-sister taxa.	A significant excess of shared derived alleles between a specific non-sister taxon pair.
Phylogenetic Network Inference	Uses methods that model both the species tree and reticulate events (hybridization) [5].	The data is best explained by a tree-like topology without reticulations.	The data supports a network topology with one or more hybridization events.
Gene Tree Discordance Distribution	Examines the distribution and support of conflicting topologies across the genome [5] [36].	Discordance is more diffuse and not strongly concentrated on a single alternative topology.	Discordance is concentrated on a specific, well-supported alternative topology involving the hybridizing taxa.
Branch Length Analysis	Looks at the length of internal branches in the species tree [5].	Short internal branches (a "rapid radiation") are conducive to ILS.	Introgression can occur regardless of internal branch lengths, though it may be easier to detect outside rapid radiations.

FAQ 3: My coalescent analysis shows high conflict at a node. How can I test if this is a "true" biological polytomy versus an artifact?

A "true" polytomy represents a hard multifurcation, often indicative of a rapid radiation where multiple lineages diverged simultaneously. An "artificial" polytomy can arise from insufficient data or methodological errors. To test this:

• Increase Data: Incorporate more loci or more informative sites (e.g., by adding taxa). A true polytomy will persist, while an artificial one may resolve [35].
• Examine Branch Support: Use methods like Overall Success of Resolution (OSR) to quantify congruence among gene trees. A true polytomy is expected to have low support for any specific resolution and a high degree of conflict that is not reduced by improving gene tree estimation [37].
• Apply Polytomy Tests: Implement statistical tests, such as the likelihood-ratio test for polytomies, to evaluate whether a multifurcating tree fits the data significantly worse than a fully resolved, binary tree. A non-significant result is consistent with a true polytomy [37].
• Check for Systematic Error: Rule out artifacts like long-branch attraction or model misspecification that can create unresolved nodes by obscuring the true phylogenetic signal [35].

FAQ 4: What are the best practices for filtering or subsampling gene trees to reduce the impact of estimation error in my concordance analysis?

Gene-tree-inference error is a major source of artifact in species-tree estimation. Filtering gene trees can improve robustness.

• Collapse Weakly Supported Nodes: Before analysis, collapse branches in gene trees with low support (e.g., below a certain bootstrap or approximate-likelihood-ratio-test threshold). This creates partially resolved trees (polytomies) that more accurately represent the uncertainty in the gene tree estimate [37].
• Subsample by Congruence: After collapsing weak branches, subsample gene trees based on their pairwise congruence with others. Use a metric like Overall Success of Resolution (OSR), which counts both matching and contradicting clades without penalizing polytomies. Exclude the most incongruent gene trees, as they are likely to contain the highest error [37].
• Recommended Workflow: A highly effective approach is to collapse likelihood-based gene-tree branches with 0% SH-like aLRT support and then subsample the resulting trees using the OSR congruence measure. This has been shown to improve coalescent branch lengths and topological robustness [37].

Troubleshooting Guides

Problem 1: Your species tree analysis is dominated by widespread gene tree conflict, and you cannot identify the primary source.

• Potential Cause: A complex mixture of biological processes (ILS, introgression, paralogy) and systematic errors.
• Solution: A Step-by-Step Diagnostic Protocol
  • Quantify and Localize Conflict: Use software like phyparts to calculate the number of gene trees that support (are concordant with) versus conflict with each node in your species tree. This will map the landscape of discordance [36].
  • Test for Introgression: Apply site pattern tests (e.g., D-statistics) to the genomic alignment to detect signals of gene flow between specific taxon pairs [5].
  • Assess the Role of ILS: Estimate the branch lengths of your species tree. Very short internal branches are a hallmark of conditions ripe for ILS. Use coalescent theory to calculate the expected amount of discordance under ILS alone and compare it to your observed distribution [5].
  • Check for Paralogy: Scrutinize your homology assessments. Ensure your gene trees are built from orthologs, not paralogs. High levels of gene duplication within clades can be a sign of hidden paralogy [36].
  • Evaluate Data Quality: Investigate potential systematic errors. Test for compositional heterogeneity and substitutional saturation in your alignments, which can cause artifactually strong conflict [35].

Diagram: A logical workflow for diagnosing the source of gene tree conflict.

Problem 2: Your concordance analysis reveals a specific node with both high conflict and low support, suggesting a potential polytomy.

• Potential Cause: A "true" hard polytomy from a rapid radiation, or an "artificial" polytomy caused by uninformative gene trees or model violation.
• Solution: A Protocol for Polytomy Testing
  • Filter Gene Trees: Apply the recommended filtering for gene tree error. Collapse low-support branches (e.g., <0% aLRT) and subsample based on OSR congruence [37].
  • Re-run Coalescent Analysis: Re-estimate the species tree with the filtered set of gene trees. Observe if the node remains unresolved.
  • Conduct a Statistical Test: Perform a likelihood-ratio test or a similar statistical framework to compare the fit of a resolution at the node versus a polytomy. A non-significant p-value suggests the polytomy is not worse than a binary resolution [37].
  • Check for Anomalous Zones: Determine if your tree falls within an "anomaly zone," where the most likely gene tree topology differs from the species tree topology due to extreme ILS. This can create the illusion of a polytomy even with a bifurcating history [5].
  • Report as Polytomy: If the evidence consistently points towards a lack of resolution, it is methodologically sound to represent the node as a polytomy in your final hypothesis, as this honestly represents the uncertainty in the evolutionary history [5] [37].

The Scientist's Toolkit: Research Reagent Solutions

Table 2: Essential Software and Analytical Tools for Concordance Analysis

Tool / Resource	Primary Function	Application in Conflict Analysis
phyparts [36]	Calculates concordance and conflict between gene trees and a species tree.	Maps gene tree discordance across the phylogeny, identifying nodes with significant conflict and calculating metrics like Internode Certainty (IC).
OSR / CongSort [37]	Quantifies topological congruence among gene trees, including those with polytomies.	Used to subsample gene trees based on their pairwise congruence, effectively filtering out erroneous trees to improve downstream coalescent analyses.
D-statistics [5]	A site-pattern based test for detecting introgression.	Provides a statistical test to distinguish introgression from ILS by identifying an excess of shared derived alleles between non-sister taxa.
PhyloNet [5]	Infers phylogenetic networks.	Models evolutionary histories that include both divergence (tree-like) and hybridization/introgression (reticulate) events.
RAxML [36]	Performs maximum likelihood phylogenetic inference.	Used for estimating individual gene trees; branches with low support (e.g., from SH-like aLRT) can be collapsed to represent uncertainty.
BUCKy [36]	Estimates the primary concordance tree from multi-locus data.	Infers the tree that represents the most common phylogenetic history across genes, directly accounting for discordance.

Frequently Asked Questions (FAQs)

1. What are the primary causes of gene tree discordance I might encounter in my phylogenomic data? Gene tree discordance, where gene trees differ from each other and from the species tree, is common in phylogenomic studies. The main causes are:

• Incomplete Lineage Sorting (ILS): The failure of genetic lineages to coalesce (find a common ancestor) within the time span between successive speciation events. This stochastic process is common in rapid radiations and leads to the retention of ancestral polymorphisms [38] [35].
• Introgression/Hybridization: The transfer of genetic material between distinct species or lineages through hybridization and backcrossing. This can lead to specific genomic regions having a different history from the rest of the genome [38] [39] [40].
• Other Factors: Gene duplication and loss, hidden paralogy, and horizontal gene transfer can also cause discordance. Furthermore, systematic errors, such as Long Branch Attraction (LBA) caused by model misspecification or heterogeneous evolutionary rates, can create artificial discordance [35].

2. How can I determine if the discordance in my dataset is due to ILS or introgression? Distinguishing between ILS and introgression requires specific tests and analyses, as their signals can be similar.

• Use D-statistics (ABBA-BABA tests): This is a common test to detect signals of introgression between two closely related species relative to a third. A significant D-statistic indicates an excess of shared derived alleles, which is consistent with introgression [39].
• Apply the QuIBL Method: The Quantifying Introgression via Branch Lengths (QuIBL) method helps distinguish ILS from introgression by analyzing the distribution of internal branch lengths for a given three-taxon subtree. ILS alone produces an exponential distribution of branch lengths for discordant topologies, while introgression adds a second component with a non-zero mode, corresponding to the time of the introgression event [39].
• Infer Phylogenetic Networks: Tools like PhyloNet can infer evolutionary networks that explicitly model reticulate events (e.g., hybridization, introgression) alongside the species tree. A network that fits your data significantly better than a tree can be strong evidence for introgression [39] [41].

3. My chloroplast and nuclear DNA phylogenies are conflicting. What does this mean and how should I proceed? This is known as cytonuclear discordance and is a frequent occurrence in plant phylogenetics [42].

• Causes: Chloroplasts are often maternally inherited and form a single genetic linkage group, making their history susceptible to processes like "chloroplast capture" via ancient hybridization and introgression. The nuclear genome, being biparentially inherited, may tell a different story [42].
• Solution: An integrative workflow is recommended. Use multiple sequence alignments from many nuclear genes to infer a species tree and then compare it to the chloroplast tree. Methods that analyze gene tree concordance (e.g., using tools like ASTRAL) and test for introgression (e.g., D-statistics) can then determine if ILS, recent introgression, or ancient introgression is the primary cause of the discordance [42].

4. What are the best practices for designing a phylogenomic study to minimize systematic error? Systematic error is a major challenge in phylogenomics, where model violation leads to strongly supported but incorrect topologies [35].

• Use Complex Models: Employ models of sequence evolution that account for heterogeneity across sites and lineages, such as partition models or profile mixture models [35].
• Improve Taxon Sampling: A denser sampling of taxa can break up long branches, which are a primary cause of Long Branch Attraction (LBA) [35].
• Critically Evaluate Data Properties: Check your data for compositional heterogeneity and saturation, which can mislead phylogenetic inference. Using amino acid sequences instead of nucleotides can sometimes reduce these effects [35].

Troubleshooting Guides

Issue 1: Widespread Gene Tree Discordance with No Clear Pattern

Problem: Your analysis of thousands of nuclear genes reveals extensive conflict among gene trees, and no single topology is highly predominant.

Diagnosis: This pattern, where the most common gene tree topology is found in only a small percentage of genomic windows (e.g., 4.3% as seen in Heliconius butterflies), is a classic signature of a complex evolutionary history involving both Incomplete Lineage Sorting (ILS) and widespread introgression [39].

Solution Steps:

• Reconstruct a Consensus Species Tree: Use a coalescent-based method (e.g., ASTRAL) that is robust to ILS to infer a primary species tree from your set of gene trees [38] [42].
• Test for Genome-Wide Introgression: Calculate D-statistics for multiple triplets of taxa to identify which lineages show significant signals of introgression [39].
• Quantity the Contribution of Introgression: Apply the QuIBL method to triplets of interest. QuIBL can estimate the proportion of discordant loci that are due to introgression versus ILS. For example, in one study, QuIBL inferred that 71% of loci with discordant gene trees had a history of introgression [39].
• Model Reticulate Evolution: Use a Bayesian network inference tool like SnappNet (in BEAST 2) or methods in PhyloNet to co-estimate the species network and parameters such as inheritance probabilities. This provides a model that explicitly includes hybridization events [41].

Relevant Experimental Protocol: Distinguishing ILS from Introgression with QuIBL

• Input Data: A genome alignment and a hypothesized species tree for a triplet of taxa (P1, P2, P3) [39].
• Method: a. Divide the genome into many small, non-recombining windows (e.g., 5 kb). b. For each window, infer a gene tree and its internal branch length. c. Plot the distribution of internal branch lengths for all gene trees that support a topology discordant from the species tree. d. Model Fitting: Fit two models to the distribution of discordant branch lengths: * ILS-only model: A single exponential distribution. * Introgression model: A mixture of an exponential (for ILS) and a gamma distribution (for introgressed loci). e. Statistical Test: Use a criterion like BIC (Bayesian Information Criterion) to determine which model best fits the data. A significant improvement for the introgression model (e.g., ΔBIC > 10) provides evidence for introgression [39].

Diagram 1: The QuIBL analysis workflow for distinguishing ILS from introgression.

Issue 2: Strong Cytonuclear Discordance in Plant Phylogenetics

Problem: The phylogenetic tree inferred from whole chloroplast genomes conflicts with the tree inferred from hundreds of nuclear genes.

Diagnosis: This is a strong indicator of complex evolutionary events. The chloroplast history may not represent the species history due to chloroplast capture, a form of ancient introgression where the chloroplast of one species is transferred into the nuclear background of another via hybridization [42].

Solution Steps:

• Assemble Both Datasets from the Same Reads: Use a tool like Read2Tree to assemble sequences for conserved nuclear genes directly from raw whole-genome sequencing reads. The same reads can be used to assemble the complete chloroplast genome, ensuring data consistency [42].
• Infer Robust Nuclear Phylogeny: Use the concatenated nuclear alignment from Read2Tree, along with coalescent-based methods on individual gene trees, to establish the most likely species tree [42].
• Analyze Gene Tree Concordance: Use the multiple sequence alignments for each orthologous group (OG) generated by Read2Tree to infer individual nuclear gene trees. Calculate the frequency of different topological signals to quantify discordance [42].
• Test for Introgression: Apply D-statistics to both the nuclear and chloroplast data to test for signals of introgression that could explain the discordant placement of specific taxa [42].

Issue 3: Inconsistent Phylogenetic Results Across Different Genomic Regions

Problem: The inferred phylogeny changes depending on which genomic region or chromosome you analyze.

Diagnosis: This heterogeneity is often correlated with genome architecture. Studies have shown that introgression is more common in genomic regions of high recombination and low gene density, as these regions are less constrained by linked selection that would remove introgressed alleles due to genetic incompatibilities [39].

Solution Steps:

• Map Topologies to the Genome: Conduct a sliding-window analysis across chromosomes to identify which phylogenetic topology is supported in each window [39].
• Correlate with Genomic Features: Calculate the correlation between the frequency of a particular introgressed topology and features like chromosome length (inversely proportional to recombination rate per base pair), local recombination rate, and gene density. A strong negative correlation with chromosome size is a tell-tale sign of selection against introgressed regions on larger (low-recombination) chromosomes [39].
• Inspect Specific Loci: Investigate long, uninterrupted blocks supporting a minor topology, as these may be caused by the introgression of large structural variants, such as inversions, which suppress recombination [39].

Table 1: Key Metrics from Phylogenomic Studies on Introgression and ILS

Study System	Analysis Method	Key Quantitative Finding	Interpretation
Heliconius Butterflies [39]	QuIBL	On average, 71% of loci with discordant gene trees were due to introgression.	Introgression, not ILS, was the dominant cause of genealogical discordance in this adaptive radiation.
Heliconius Butterflies [39]	Topology Frequency	The most common gene tree topology was found in only 4.3% of genomic windows.	Phylogenetic discordance was widespread across the genome, with no single dominant history.
Heliconius Butterflies [39]	Chromosomal Correlation	Tree 1 (introgressed) frequency vs. chromosome length: r² = 0.883.	Strong evidence that introgressed regions are purged more efficiently on longer (low-recombination) chromosomes.
Oleaceae [38]	Network Analysis	Tribe Oleeae originated via ancient hybridization, with one parent being a "ghost lineage."	Phylogenetic conflict at deep timescales can be explained by hybridization events that are no longer visible in extant diversity.

Table 2: Comparison of Model-Based Network Inference Tools

Tool / Method	Input Data	Core Model	Key Features	Considerations
SnappNet [41]	Biallelic markers (e.g., SNPs)	Multispecies Network Coalescent (MSNC)	Bayesian; integrates over all possible gene trees; fast likelihood computation.	Implemented in BEAST 2; efficient for larger datasets.
PhyloNet (MCMC_BiMarkers) [41]	Biallelic markers (e.g., SNPs)	Multispecies Network Coalescent (MSNC)	Bayesian; jointly samples networks and gene trees.	Can be computationally intensive on complex networks.
PhyloNet (Inference from Gene Trees) [41]	Pre-inferred Gene Trees	Multispecies Network Coalescent (MSNC)	Maximum Likelihood or Bayesian; uses gene trees as input.	Faster than full-data methods, but may lose signal in sequence alignments.

The Scientist's Toolkit: Essential Research Reagents & Solutions

Table 3: Key Software and Analytical Tools for Complex Phylogenomics

Tool / Resource	Function	Application in Distinguishing ILS/Introgression
Read2Tree [42]	Assembles conserved nuclear genes and constructs phylogenies from raw sequencing reads.	Cost-effective generation of nuclear phylogenies from the same data used for chloroplast assembly, enabling cytonuclear discordance studies.
PhyloNet [39] [41]	A software package for inferring phylogenetic networks.	Infers evolutionary networks to model hybridization and introgression events directly. Includes implementations for D-statistics and MSNC.
SnappNet [41]	A Bayesian method for phylogenetic network inference from biallelic markers.	Co-estimates species networks and parameters under the MSNC model, accounting for both ILS and introgression.
QuIBL [39]	A statistical test to distinguish ILS from introgression using branch lengths.	Quantifies the proportion of discordant gene trees caused by introgression versus ILS for specific taxon triplets.
D-Statistics (ABBA-BABA) [39]	A test for detecting allele sharing excess indicative of introgression.	Provides a genome-wide test for introgression between specific pairs of taxa relative to an outgroup.

Diagram 2: An integrated workflow for diagnosing the causes of phylogenetic discordance.

Navigating Analytical Challenges: Best Practices for Complex Datasets

Addressing Methodological Limitations and Statistical Power Concerns

Troubleshooting Guides

Guide 1: Resolving Phylogenomic Incongruence

Q: My phylogenetic analyses of different genomic datasets (e.g., nuclear vs. plastid) yield conflicting trees. How can I determine if this is due to biological reality or methodological error?

A: Incongruence between phylogenetic trees can stem from biological processes like Incomplete Lineage Sorting (ILS) and introgression, or from methodological artifacts [43]. Follow this diagnostic workflow to systematically identify the source:

Diagnostic Steps:

• First, rule out methodological artifacts [43]:

  • Test for compositional heterogeneity: Use statistical tests like χ² to check if sequence composition is homogeneous across taxa
  • Assess branch length heterogeneity: Identify long branches that might cause attraction artifacts
  • Check for site saturation: Determine if multiple substitutions have obscured phylogenetic signal
  • Verify orthology assignment: Ensure sequences are truly orthologous, not paralogous or contaminated

• If methodological issues are minimized, test biological hypotheses:

  • Calculate gene tree concordance: Quantify the proportion of gene trees supporting different topologies [6]
  • Apply coalescent-based methods: Use ASTRAL or similar tools to account for ILS
  • Test for introgression: Use D-statistics and phylogenetic networks to detect gene flow [7]

Expected Outcomes:

• High gene tree discordance with consistent asymmetry suggests ILS [6]
• Significant D-statistics with specific phylogenetic patterns indicates introgression [7]
• Model violations that resolve with better models point to methodological issues [43]

Guide 2: Addressing Statistical Power Limitations

Q: I suspect my dataset lacks power to distinguish between ILS and introgression. How can I improve my analysis?

A: Statistical power in phylogenomics depends on the number of loci, informative sites, and appropriate model selection. Address these concerns proactively:

Implementation Framework:

Table: Statistical Power Considerations for Phylogenomic Analyses

Factor	Minimum Recommendation	Ideal Scenario	Detection Improvement
Number of Loci	100-500 loci [7]	1,000+ loci [6]	Increases resolution of gene tree distributions
Taxon Sampling	1-2 representatives per clade	Multiple representatives per clade [7]	Helps distinguish ILS from introgression patterns
Informative Sites	10,000+ sites	100,000+ sites	Improves branch support and test sensitivity
Model Complexity	Partitioned models	Site-heterogeneous models (e.g., CAT) [43]	Reduces systematic error and false positives

Protocol: Power-Enhanced Hypothesis Testing

• Generate multiple dataset compositions:

  • Create subsets with varying taxon sampling
  • Test with different locus numbers
  • Analyze with increasingly complex models

• Apply consistent testing framework:

  • Calculate D-statistics for all dataset compositions
  • Perform quartet concordance analysis
  • Compare results across sampling regimes

• Assess robustness:

  • Look for consistent signal across analyses
  • Use jackknifing or bootstrapping to measure support
  • Report how conclusions change with different data treatments

Frequently Asked Questions

FAQ 1: Methodological Limitations

Q: What are the most common methodological artifacts that mimic ILS or introgression signals?

A: The most prevalent artifacts arise from model violation and data misassignment [43]:

• Branch length heterogeneity: Causes long-branch attraction that can mimic rapid diversification patterns
• Compositional heterogeneity: Creates false groupings based on similar base composition rather than shared ancestry
• Site saturation: Leads to underestimation of divergence and incorrect tree topologies
• Orthology misassignment: Results in mixing paralogous sequences that don't reflect species history

Solution: Always conduct model adequacy tests and data quality assessments before interpreting biological patterns [43].

Q: How can I determine if my evolutionary model is adequate for distinguishing ILS from introgression?

A: Use posterior predictive simulations and model comparison frameworks:

• Compare site-specific likelihood scores between models
• Test whether more complex models significantly improve fit
• Use simulations to verify that your analysis can distinguish ILS from introgression under realistic conditions
• Apply cross-validation to assess model performance

FAQ 2: Analytical Approaches

Q: What analytical methods are most effective for distinguishing ILS from introgression in genome-scale data?

A: A hierarchical approach combining multiple methods is most effective:

Table: Comparative Analysis of ILS vs. Introgression Detection Methods

Method Type	Specific Tools	Strengths	Limitations	Data Requirements
Summary Statistics	D-statistics (ABBA-BABA) [6] [7]	Simple, computationally efficient, works with genome-wide data	Cannot detect direction of introgression, sensitive to taxon sampling	Genome-wide SNP data or sequence alignments
Coalescent-Based	ASTRAL, MP-EST	Accounts for ILS, provides species tree estimates	Computationally intensive, assumes no gene flow	Multiple gene trees or alignments
Phylogenetic Networks	PhyloNet, SplitsTree	Visualizes conflicting signals, models reticulation	Complex interpretation, computationally demanding	Gene trees or sequence alignments
SINE/LTR Analysis	Presence/absence patterning [6]	Nearly homoplasy-free, clear interpretation	Limited to taxa with available mobile elements	Whole genome sequences

Q: How many genomic markers do I need to reliably distinguish ILS from introgression?

A: The required number depends on divergence time and extent of gene flow:

• For recent radiations (≤ 1 million years): 1,000+ loci are often necessary [6]
• For deeper divergences: 100-500 loci may suffice if they contain sufficient phylogenetic signal [7]
• For SINE-based analyses: Hundreds to thousands of insertion events provide strong evidence [6]

Critical consideration: More important than the absolute number is the information content of each locus. Focus on obtaining loci with sufficient length and variation rather than maximizing count alone.

Research Reagent Solutions

Table: Essential Materials and Tools for Phylogenomic Conflict Analysis

Reagent/Resource	Function	Application Context	Implementation Notes
Transcriptome Data	Provides thousands of nuclear loci for analysis [7]	Alternative to whole genomes when dealing with large genomes	Requires careful orthology assessment
SINE/LTR Markers	Nearly homoplasy-free phylogenetic characters [6]	Determining species relationships despite gene tree discordance	Limited to taxa with characterized mobile elements
Ultraconserved Elements (UCEs)	Targeted sequencing of conserved genomic regions	Phylogenetics across diverse taxonomic scales	Proven effective in Myotis phylogenetics [6]
Model Testing Software (ModelTest-NG, Modelfinder)	Selects best-fit evolutionary models [43]	Preventing model misspecification artifacts	Should be applied to each partition
Coalescent Analysis Packages (ASTRAL, SVDquartets)	Species tree inference accounting for ILS	Resolving relationships in rapidly diversifying groups	Requires well-resolved gene trees
Introgression Tests (Dsuite, PhyloNet)	Detects and quantifies gene flow	Distinguishing introgression from ILS	Complementary approaches provide validation

Optimizing Taxon and Locus Sampling to Break the ILS-Introgression Deadlock

Frequently Asked Questions

1. What are the most common causes of gene tree discordance that can confound my analysis? The two primary biological processes causing gene tree discordance are Incomplete Lineage Sorting (ILS) and introgression. ILS occurs when ancestral genetic polymorphisms persist and fail to coalesce in the immediate ancestral population, leading to gene trees that differ from the species tree. Introgression (or hybridization) involves the transfer of genetic material between species through hybridization and backcrossing. Both processes can produce similar patterns of genealogical discordance, making them difficult to distinguish without proper sampling and modeling [1]. Additional technical sources of discordance include gene tree estimation errors, particularly at deeper evolutionary timescales [1].

2. My phylogenomic analysis has strong support, but I suspect it might be wrong. What could be happening? You may be experiencing the effects of systematic error. Unlike stochastic error (which is reduced by adding more data), systematic error arises from incorrect model assumptions and can be exacerbated by larger datasets. Common causes include:

• Long-Branch Attraction (LBA): Rapidly evolving lineages or sequences are incorrectly clustered together due to convergent molecular substitutions [35] [44].
• Model misspecification: Using overly simple models of sequence evolution that fail to account for heterogeneity in substitution patterns across sites, genes, or lineages [35] [44].
• Compositional heterogeneity: Differences in nucleotide or amino acid composition between taxa that can create artifactual groupings [35].

3. How can I improve my ability to distinguish between ILS and introgression? A multi-faceted approach is most effective:

• Increase taxon sampling: Denser sampling across the phylogeny, especially by breaking up long branches, can dramatically improve accuracy and reduce systematic biases like LBA [35] [45].
• Use more complex models: Employ models that account for site-specific and lineage-specific rate variation, such as profile mixture models (e.g., the CAT model) [35] [45].
• Leverage full-likelihood methods: Methods like BPP that use multilocus sequence alignments directly (considering both gene-tree topologies and branch lengths) generally outperform heuristic approaches for detecting complex introgression scenarios, including ghost introgression [46].

4. What is "ghost introgression" and why is it particularly challenging to detect? Ghost introgression refers to gene flow from an extinct or unsampled lineage into a sampled species. It is challenging because heuristic phylogenetic methods (e.g., those based solely on site-pattern counts or gene-tree topologies, like HyDe or some PhyloNet applications) often cannot distinguish it from introgression between sampled non-sister species. These methods may incorrectly identify the donor and recipient species [46]. Full-likelihood methods are better suited for this task [46].

5. Does adding more genes or more taxa have a greater impact on breaking the deadlock? While increasing the number of loci reduces stochastic error, adding more taxa is often more critical for mitigating systematic errors like long-branch attraction. Denser taxon sampling helps by breaking long branches, providing more information about the sequence of divergence events, and allowing for better model parameter estimation [35] [45]. The most robust studies aim to maximize both, but if forced to choose, prioritizing comprehensive taxon sampling is often advisable for resolving deep divergences.

Experimental Protocols & Methodologies

Protocol 1: Designing a Phylogenomic Target Capture Study

This protocol outlines a common approach for generating phylogenomic datasets by focusing sequencing effort on pre-selected loci [47].

• Define Research Question and Taxonomic Scope: Identify the phylogenetic depth of your question (shallow vs. deep divergences) as this influences bait design and taxon sampling.
• Evaluate Available Bait Sets: Determine if a pre-designed bait set (e.g., for Ultraconserved Elements or Anchored Hybrid Enrichment) is suitable for your taxonomic group. Using existing baits saves cost and increases comparability across studies [47].
• Design Custom Baits (if necessary): If no suitable bait set exists, design custom RNA baits by:
  • Identifying target loci from genomic alignments of related species.
  • Selecting regions that are sufficiently conserved for bait hybridization but flanked by variable regions for phylogenetic informativeness.
• Sample Preparation and DNA Extraction: Extract DNA from samples. This method is suitable for a range of sample qualities, including degraded DNA from museum specimens [47].
• Library Preparation and Hybridization: Prepare sequencing libraries, hybridize them with the bait sequences, wash away non-bound DNA, and then amplify and sequence the captured targets.
• Bioinformatic Processing: Process raw sequence data through a pipeline involving quality trimming, assembly of targeted loci (contigs), and alignment to generate the final phylogenomic matrix.

Protocol 2: Detecting Introgression between Sister Species using RNDmin

This protocol uses the RNDmin statistic, a powerful and robust method for identifying introgressed genomic regions between two sister species, especially when introgression is recent or rare [48].

• Data Collection: Obtain whole-genome, phased haplotype data from two sister species and an outgroup. The outgroup should not have experienced introgression with the ingroup.
• Calculate Minimum Sequence Distance (dmin): For each locus, compute dmin, which is the minimum number of sequence differences between any haplotype from species X and any haplotype from species Y [48].
• Calculate Average Distance to Outgroup (dout): Compute the average sequence distance from each species to the outgroup: dout = (dXO + dYO)/2 [48].
• Compute RNDmin: Calculate the RNDmin statistic for each locus using the formula: RNDmin = dmin / dout [48].
• Generate a Null Distribution and Identify Outliers: Simulate a null distribution of RNDmin under a model of no migration using coalescent simulations. Genomic regions with observed RNDmin values in the lower tail of this distribution (significantly lower than expected) are strong candidates for containing introgressed sequences [48].

Method Comparison and Selection Guide

The table below summarizes key methods for detecting introgression, highlighting their uses and limitations.

Table 1: Comparison of Methods for Detecting Introgression in Phylogenomics

Method	Type	Data Input	Primary Use	Key Strengths	Key Limitations
D-statistic (ABBA-BABA)	Heuristic	Site patterns (quartets)	Tests for introgression between non-sister species [1] [46].	Simple, fast, widely used; good for initial screening [1].	Cannot identify introgressed regions; confounded by ghost introgression [46].
HyDe	Heuristic	Site patterns (quartets)	Identifies hybrid species and introgression [46].	Based on a hybrid speciation model.	Can misidentify donor/recipient in outflow and ghost introgression scenarios [46].
PhyloNet/MPL	Heuristic (Pseudo-likelihood)	Gene trees	Infers phylogenetic networks from multi-taxon data [46].	Useful for visualizing complex relationships.	Relies only on gene-tree topologies; network identifiability can be an issue [46].
RNDmin	Summary Statistic	Phased haplotypes (trios)	Detects introgressed regions between sister species [48].	Robust to mutation rate variation; sensitive to recent and rare introgression [48].	Requires phased data and an outgroup; power depends on introgression timing/strength [48].
BPP	Full-Likelihood	Multi-locus sequence alignments	Co-estimates species trees, divergence times, and introgression under the MSC model [46].	Uses all information (topologies & branch lengths); powerful for detecting ghost introgression [46].	Computationally intensive [46].

The Scientist's Toolkit: Research Reagent Solutions

Table 2: Essential Materials and Resources for Phylogenomic Studies

Item / Resource	Function / Application	Notes
Ultraconserved Elements (UCE) Probe Sets	Target capture bait sets for enriching conserved genomic regions across divergent taxa [47].	Probes are designed for conserved "core" regions with variable flanks; allow consistent locus sampling across deep evolutionary scales [47].
Anchored Hybrid Enrichment (AHE) Probe Sets	Target capture bait sets designed to target conserved exonic regions flanked by more variable segments [47].	Similar goal to UCEs; another common approach for phylogenomics [47].
BPP (Bayesian Phylogenetics and Phylogeography)	Software for Bayesian inference of species trees, population parameters, and introgression from multilocus sequence data [46].	A full-likelihood method that is particularly effective for testing complex introgression scenarios, including ghost introgression [46].
PhyloNet	Software package for inferring and analyzing phylogenetic networks [46].	Contains tools like InferNetwork_MPL which use gene tree topologies to infer networks [46].
Profile Mixture Models (e.g., CAT model)	Complex models of sequence evolution that account for heterogeneity in amino acid preferences across sites [35] [45].	Can reduce systematic errors like long-branch attraction; computationally demanding but more biologically realistic [45].

Workflow and Conceptual Diagrams

Phylogenomic Analysis Workflow

Distinguishing ILS from Introgression

Interpreting Inheritance Probabilities (γ) in Phylogenetic Networks

Frequently Asked Questions (FAQs)

Q1: What does the inheritance probability (γ) represent in a phylogenetic network?

In a phylogenetic network, an inheritance probability (γ), also known as a hybridization parameter, is assigned to hybrid edges. It quantifies the proportional genetic contribution from a specific parental lineage to a hybrid descendant. These parameters are defined for each hybrid edge, with the sum of γ values for all edges leading into the same hybrid node equaling 1. For tree edges, the inheritance probability is always 1, as they represent direct vertical descent [49].

Q2: How can I distinguish if a gene tree conflict is caused by ILS or introgression?

Distinguishing between Incomplete Lineage Sorting (ILS) and introgression is a central challenge, as both processes can produce similar patterns of gene tree conflict [50]. However, their underlying mechanisms differ [50]:

• ILS occurs when ancestral genetic polymorphisms fail to coalesce (merge to a common ancestor) in the time between successive speciation events. It is a neutral process more common in rapidly diverging lineages with large effective population sizes [50] [51].
• Introgression involves the transfer of genetic material from one species into the gene pool of another through hybridization and backcrossing. It can introduce alleles with evolutionary histories that differ from the species tree and may be adaptive [50] [52].

The table below summarizes the key characteristics to help differentiate them.

Table: Differentiating Between ILS and Introgression

Feature	Incomplete Lineage Sorting (ILS)	Introgression
Underlying Process	Retention of ancestral polymorphisms [50].	Exchange of genetic material post-divergence [50].
Genomic Signature	Randomly distributed gene tree conflicts across the genome [50].	Localized gene tree conflicts in specific genomic regions [50] [51].
Linkage to Adaptation	Typically neutral [50].	Can be neutral or adaptive [52] [51].
Common Detection Methods	Tests for ancestral polymorphism, gene tree concordance analysis [50].	D-statistic (ABBA-BABA), Phylonetwork analyses, branch-length tests [50] [49].

Q3: My analysis detected significant introgression, but the γ value for the hybrid branch is very low (<0.1). How should I interpret this?

A low γ value indicates a minor genetic contribution from one parental lineage. Biologically, this could represent several scenarios:

• Ancient Admixture: A single, historically minor hybridization event.
• Continuous Gene Flow: Persistent but limited gene flow, where only small portions of the genome introgress.
• Adaptive Introgression: The selective introgression of a few beneficial alleles, which does not require a large genomic contribution.

From a methodological perspective, accurately estimating very low γ values can be challenging. It is essential to verify that the signal is statistically robust and not an artifact of model misspecification or insufficient data. A low γ value does not diminish the biological importance of the introgression, especially if the introgressed regions are functionally significant [49].

Q4: What are the best practices for testing the robustness of my γ estimates?

To ensure the reliability of your inheritance probability estimates, follow these practices:

• Data Resampling: Perform bootstrap analyses on your gene trees or sequence alignments to assess the variance and confidence intervals of your γ estimates.
• Model Comparison: Use statistical frameworks like Approximate Bayesian Computation (ABC) to test different evolutionary scenarios and compare model fit [50].
• Data Type Sensitivity: Analyze your data with different methods (e.g., summary statistics like D-statistics vs. full probabilistic models) to see if γ estimates are consistent [52] [49].
• Parameter Identifiability: Be aware that in certain network structures, particularly more complex blobs, some parameters may not be fully identifiable from the data alone, even with infinite data [49].

Troubleshooting Guides

Problem: Inconsistent Phylogenetic Signals Across Genomic Regions

Symptom: Different genomic regions or gene trees support strongly conflicting phylogenetic relationships, and you are unsure if this is due to ILS, introgression, or other factors.

Step-by-Step Diagnostic Protocol:

• Quantify the Incongruence

  • Action: Calculate quartet concordance factors (CFs) for all subsets of four taxa. CFs are the proportions of gene trees that display each of the three possible unrooted quartet topologies [49].
  • Tool Suggestion: Use software like ASTRAL or the underlying methods in SNaQ/NANUQ [49].

• Test for Introgression

  • Action: Perform a D-statistic (ABBA-BABA) test to detect a significant excess of shared derived alleles between non-sister taxa, which is a signature of introgression [50].
  • Interpretation: A significant D-statistic is evidence of introgression. The sign of the statistic can indicate the direction of gene flow [50].

• Interrogate Gene Genealogies

  • Action: Use Gene Genealogy Interrogation (GGI) to compare the distribution of gene tree topologies to the expected distribution under a pure coalescent model with ILS [50].
  • Interpretation: An excess of gene trees supporting a specific alternative topology beyond the expectations of ILS suggests introgression [50].

• Scan for Genomic Islands

  • Action: Calculate genetic differentiation (e.g., F~ST~) or divergence metrics in sliding windows across the genome. Look for "islands" of elevated differentiation.
  • Interpretation:
    • Islands primarily shaped by linked selection are expected to correlate with regions of low recombination and may be shared across lineages if they are old [51].
    • Islands that are lineage-specific and contain tracts of shared ancestry from introgression are likely shaped by adaptive gene flow. Correlate these regions with functional genomic data [51].

The following diagram illustrates this multi-step diagnostic workflow.

Problem: Weak Statistical Support for Inferred Reticulate Nodes

Symptom: The phylogenetic network inference algorithm identifies a hybrid node, but the statistical support (e.g., via bootstrap) for that node is low.

Potential Causes and Solutions:

Table: Troubleshooting Weak Support for Reticulations

Potential Cause	Diagnostic Check	Proposed Solution
Insufficient Data	Check the number of informative sites or genes.	Increase sequencing depth or the number of sampled loci. Transcriptome or whole-genome data is often necessary [50].
Weak Signal	The true γ value may be very close to 0.5 (equal contribution) or very low (<0.1), making it hard to distinguish from a tree.	Use methods specifically designed to detect minor introgression. Acknowledge the uncertainty in interpretation [49].
Model Violation	The evolutionary model used does not account for key features of the data (e.g., rate variation, selection).	Employ model testing. Use methods that incorporate population-level parameters or are robust to some model violations [49].
Network Identifiability Issue	The network structure itself might be non-identifiable from the data type used.	Consult theoretical work on identifiability. For quartet CFs, level-1 networks and certain galled tree-child networks are known to be identifiable [49].

The Scientist's Toolkit: Research Reagent Solutions

Table: Essential Materials and Tools for Phylogenomic Network Analysis

Item / Reagent	Function / Explanation	Example Tools / Uses
High-Quality Genomic DNA	Starting material for whole-genome resequencing. Essential for obtaining a genome-wide, unbiased set of markers.	Used in population genomics studies of recent radiations [51].
RNA Extraction Kits	To obtain transcriptomic data from fresh tissue. Useful for phylogenetic studies by targeting conserved, expressed genes.	Used in phylogenetic studies like the Aspidistra analysis [50].
PCR and Sequencing Reagents	For generating data from specific loci (e.g., Sanger sequencing) or for library preparation for NGS.	Amplifying specific genes or preparing libraries for RAD-seq, ultra-conserved elements.
Bioinformatics Pipelines	Software for processing raw sequencing data into analyzable formats.	Variant callers (GATK, SAMtools), alignment tools (BWA, MAFFT).
Phylogenetic Network Software	Specialized tools to infer networks and estimate parameters like γ from gene tree or sequence data.	SNaQ [49]: Infers networks from quartet concordance factors.PhyNEST [49]: Infers networks from site pattern frequencies.Phylonet [52]: Suite for inferring networks and calculating D-statistics.
Accessible Color Palettes	A set of colors with sufficient contrast for creating accessible data visualizations and figures.	Use tools like Viz Palette [53] or Coolors [54] to ensure charts are readable by those with color vision deficiencies (CVD).

The diagram below summarizes the logical relationships and workflow from data generation to network inference, highlighting key software tools.

Distinguishing Recent vs. Ancient Introgression Events

FAQ: Core Concepts and Common Challenges

1. What is the fundamental difference between signals of recent and ancient introgression in genomic data?

Recent introgression creates strong, block-like patterns of linkage disequilibrium (LD), where a long, continuous segment of DNA from the donor species is found in the recipient species' genome. Ancient introgression, however, is characterized by shorter, more fragmented introgressed segments due to millions of years of recombination breaking down the original haplotype blocks [55] [56].

2. How can we distinguish introgression from Incomplete Lineage Sorting (ILS)?

Both processes can cause gene tree-species tree discordance, but they leave different signatures. ILS produces discordance that is scattered randomly across the genome and follows a predictable probability distribution under the multispecies coalescent model. In contrast, introgression often results in a geographically restricted signal, where specific genomic regions show a stronger phylogenetic affinity to a distantly related lineage than to a closely related one [55] [57] [56]. Statistical tests like the D-statistic (ABBA-BABA test) are designed to detect this excess of shared derived alleles between non-sister taxa, which is a hallmark of introgression [55].

3. What are the major limitations when working with non-model organisms or datasets with limited sampling?

A key challenge is that a "democratic majority tree" (the species tree inferred from the most frequent gene tree topology) may not represent the true species history if ancient gene flow has affected large portions of the genome [56]. Furthermore, limited taxonomic sampling, especially the absence of lineages critical to resolving key nodes, can make it impossible to differentiate between alternative phylogenetic scenarios, such as distinguishing a hybrid origin from ILS [57] [56].

4. My D-statistic is significant. Does this confirm recent introgression?

A significant D-statistic indicates a deviation from the expected tree-like model of evolution and is often interpreted as evidence for gene flow. However, it does not, by itself, quantify the proportion or timing of introgression [55]. Follow-up analyses, such as the f-statistics (e.g., fd, fhom) or model-based approaches in tools like PhyloNet, are required to estimate the admixture proportion and to further test the timing of the introgression event [55].

Troubleshooting Guides

Guide 1: Selecting the Right Test for Your Data

Table 1: Summary of Key Methods for Detecting Introgression

Method	Core Principle	Optimal Data Requirements	Strengths	Potential Pitfalls
D-statistic (ABBA-BABA) [55]	Compares frequencies of "ABBA" and "BABA" site patterns in a 4-taxon quartet (((P1,P2),P3),O).	Four taxa; unlinked SNP data or many short loci (e.g., RAD-seq).	Fast, simple; robust to ILS when lineages are unlinked.	Only tests for presence/absence of gene flow; does not provide direction or proportion; confounded by ancestral structure.
f-branch statistic (f_d) [55]	Extends D-statistic to estimate the proportion of ancestry from a donor population.	Population-level data for P3 (two or more individuals).	Quantifies admixture proportion; useful for recent introgression.	Requires population data; performance on ancient introgression not well established.
DFOIL [55]	Extension of D-statistic logic to a 5-taxon tree, allowing inference of the direction of gene flow.	Five taxa with known topology (((P1,P2),(P3,P4)),O).	Infers directionality of introgression.	More complex; requires a fifth lineage to polarize direction.
PhyloNet [55]	Uses a likelihood framework to infer phylogenetic networks directly from gene trees or sequences.	Genome-scale multiple sequence alignments or pre-inferred gene trees.	Models introgression and ILS simultaneously; infers explicit network; can handle complex scenarios.	Computationally intensive; requires expertise in model selection.
Phylogenomic Network Analysis [56]	Visualizes conflicting phylogenetic signals across the genome as a network rather than a tree.	Genome-wide data from multiple individuals per species.	Ideal for visualizing and testing for ancient gene flow that pervasively misleads species tree inference.	Signal can be difficult to interpret, especially with multiple successive introgressions.

Guide 2: Designing an Effective Analysis Workflow

Follow the decision workflow below to choose the appropriate analytical path for distinguishing introgression from ILS.

Guide 3: Addressing False Positives and Technical Artifacts

Problem: A significant signal of introgression is detected between two species that have no known history of contact.

• Possible Cause 1: Inadequate Taxon Sampling.
  • Solution: Re-run your analysis including potential "ghost" lineages (unsampled or extinct taxa) as outgroups. Ghost lineages can create signals that mimic introgression between sampled taxa [56].
  • Example: A study on Pallas's leaf warbler found that mitochondrial DNA that appeared to introgress from another species was actually inherited from an extinct "ghost" lineage [56].
• Possible Cause 2: Model Violation.
  • Solution: Test the robustness of your result using model-based methods like PhyloNet that can jointly model ILS and introgression. Cross-validate the signal with multiple different statistical tests (e.g., D-statistic, f4-statistic) [55] [58].
• Possible Cause 3: Sequencing or Assembly Bias.
  • Solution: Check for regions of exceptionally high or low coverage that might correlate with the introgression signal. Re-map reads to a different reference genome if available, to rule out reference bias.

Problem: The inferred direction of gene flow from DFOIL analysis seems biologically implausible.

• Possible Cause: Mis-specified Species Topology.
  • Solution: DFOIL requires a known and correct 5-taxon species tree [55]. Re-assess your backbone species phylogeny using multiple, complementary methods (e.g., concatenation, ASTRAL) with high-support loci. Even small errors in the assumed topology can invalidate the direction inference.

The Scientist's Toolkit

Table 2: Essential Research Reagent Solutions for Phylogenomic Introgression Studies

Tool / Reagent	Critical Function in Analysis
High-Quality Reference Genomes	Essential for accurate read mapping, variant calling, and phasing of haplotypes to detect introgressed blocks.
Annotated Genome Assemblies	Allows for functional analysis of introgressed regions (e.g., are they enriched for genes involved in immunity or adaptation?).
Variant Call Format (VCF) Files	The standard format storing genotype calls across multiple individuals; the primary input for many population genetic tests.
Multiple Sequence Alignment (MSA) Files	Whole-genome or locus-specific alignments used for gene tree inference and model-based analyses in PhyloNet.
Outgroup Genomes	Critical for polarizing alleles as ancestral (A) or derived (B) in ABBA-BABA tests and for rooting phylogenetic trees [55].
Software: PhyloNet	Infers phylogenetic networks and explicitly quantifies introgression from gene tree data, handling both ILS and gene flow [55].
Software: Dsuite	Efficiently calculates D-statistics, f-branch statistics, and related metrics for many taxon quadruplets across the genome.
Fossil Calibration Data	Provides absolute time estimates for divergence nodes, crucial for contextualizing whether introgression is recent or ancient.

Managing Computational Demands and Data Integration Challenges

Frequently Asked Questions (FAQs)

Q1: What are the primary biological processes that cause gene tree discordance, and how can I distinguish between them?

Gene tree discordance, where gene trees from different loci show conflicting topologies, is primarily caused by two biological processes: Incomplete Lineage Sorting (ILS) and introgression. Distinguishing between them is a core challenge in phylogenomics [1].

• Incomplete Lineage Sorting (ILS) occurs when ancestral genetic variation persists through speciation events, causing some gene trees to reflect a history that differs from the species tree. Under the multispecies coalescent model, the two discordant gene tree topologies are expected to be equal in frequency. The probability of ILS is 1-e-τ, where τ is the length of the internal branch in coalescent units [1].
• Introgression (or hybridization) involves the transfer of genetic material between two species or lineages via hybridization and backcrossing. Unlike ILS, introgression often produces strongly asymmetric patterns of gene tree discordance, where one discordant topology is significantly more frequent than the other [1].

Q2: What is the minimum data requirement for powerful tests of introgression like the D-statistic?

The minimum requirement is genomic data from a rooted triplet of species (three focal species) or an unrooted quartet (three focal species plus an outgroup). This can be done using a single haploid sequence per species, with data sampled from many loci across the genome [1].

Q3: My analysis has revealed high levels of gene tree heterogeneity. What are the first steps to determine if introgression is the cause?

First, establish ILS as your null hypothesis. Calculate the frequencies of the three possible gene tree topologies for your quartet. Under a pure ILS model, the two discordant topologies are expected to be equal in frequency. A significant deviation from this symmetry, where one discordant topology is over-represented, is a key signature of introgression. Methods like the D-statistic (ABBA-BABA test) are designed to detect this asymmetry [1].

Q4: How can I characterize the direction and timing of an introgression event I have detected?

Characterizing introgression goes beyond simple detection. To infer the direction (donor and recipient populations) and timing of the event, you will need to use model-based likelihood inference methods. These include approaches for inferring phylogenetic networks, which can model introgression as instantaneous "pulses" or continuous gene flow. These methods use the distribution of gene tree topologies and branch lengths across the genome to estimate these parameters [1].

Q5: What are common pitfalls or misinterpretations when using summary statistics like the D-statistic?

A common pitfall is failing to account for other factors that can cause gene tree discordance. The D-statistic is powerful because its genealogical signal is generally not mimicked by selection, making it robust to non-neutral processes. However, it is crucial to remember that other factors, such as gene tree estimation errors (especially at older timescales), can also contribute to discordance and should be considered when interpreting results [1].

Troubleshooting Common Experimental Issues

Issue: Inconsistent or conflicting signals of introgression across different genomic regions.

• Potential Cause: This is an expected consequence of the recombination landscape. Recombination uncouples the history of neighboring genomic windows, meaning that tracts of introgressed DNA are interspersed with regions of the genome that have a different history. The extent of introgression will naturally vary across the genome [1].
• Solution: This is not necessarily an error. Analyze the distribution of introgressed signals (e.g., using fd or related statistics) to identify genomic blocks with a strong signal. This pattern can itself be informative about the history and selection on the introgressed regions.

Issue: Low statistical power to detect introgression.

• Potential Cause 1: The introgression event was very ancient, and the introgressed segments have been broken down by recombination over many generations, making them difficult to detect.
• Solution 1: Increase the number of loci analyzed. Using whole-genome data maximizes the chance of capturing loci that retain a signal of the ancient event.
• Potential Cause 2: The internal branch length (τ) of the species tree is long, reducing the expected frequency of discordant gene trees and making it harder to distinguish introgression from a small amount of ILS.
• Solution 2: Ensure you are using methods with an appropriate null model that accounts for the expected level of ILS given your species tree. Model-based approaches that co-estimate ILS and introgression parameters may be more powerful in these scenarios [1].

Issue: Gene tree estimation error is swamping the phylogenetic signal.

• Potential Cause: Gene trees inferred from individual loci, particularly those with low sequence variation or high rates of evolution, can be inaccurate. This estimation error introduces noise that can be misinterpreted as biological discordance.
• Solution: Use methods that account for gene tree uncertainty in their inference. Alternatively, use full-likelihood methods that operate directly on sequence alignments rather than relying on pre-estimated gene trees. For a large-scale analysis, ensure you are using high-quality, reliably aligned loci [1].

Key Metrics and Experimental Protocols

The table below summarizes critical metrics and thresholds used in distinguishing ILS from introgression.

Metric / Parameter	Description	Formula / Threshold	Interpretation / Use Case
D-statistic (ABBA-BABA)	Test for asymmetry in gene tree frequencies indicative of introgression [1].	D = (NABBA - NBABA) / (NABBA + NBABA)	A significant deviation from D=0 suggests introgression. Requires an outgroup.
Probability of ILS	The probability that lineages fail to coalesce in their most recent ancestral population [1].	P(ILS) = e-τ	Where τ is the internal branch length in coalescent units (2N generations).
Gene Tree Frequency Symmetry	The expected frequency of the two discordant gene tree topologies under an ILS-only model [1].	P(Tree2) = P(Tree3) = (1/3)e-τ	A significant asymmetry between these two frequencies is a signal of introgression.
Contrast Ratio (for Visualizations)	Minimum contrast for graphical objects in charts and diagrams as per WCAG 2.1 AA [59].	3:1	Ensures that graphical elements in figures (bars, lines, etc.) are distinguishable for all users.

Protocol: Distinguishing ILS from Introgression Using Quartet Sampling

Objective: To detect and characterize introgression between three ingroup species while accounting for background levels of ILS.

Materials & Computational Tools:

• Whole-genome sequencing data from a minimum of three ingroup species and one outgroup.
• Genome alignment for the target species.
• Software capable of calculating D-statistics (e.g., Dsuite) and/or inferring phylogenetic networks (e.g., PhyloNet, SNaQ).

Methodology:

• Data Preparation: Generate a whole-genome alignment for your three ingroup species (P1, P2, P3) and an outgroup (O).
• Locus Selection: Divide the genome alignment into many non-overlapping windows or extract individual loci. Ensure each locus is long enough for reliable tree estimation but short enough to assume no intra-locus recombination.
• Gene Tree Inference: Estimate a gene tree for each locus/window. This will result in a collection of thousands of gene trees.
• Calculate Gene Tree Frequencies: Count the frequency of the three possible rooted tree topologies: [((P1,P2),P3),O], [((P1,P3),P2),O], and [((P2,P3),P1),O].
• Test for Asymmetry: Apply the D-statistic or a related test to the site patterns or tree frequencies. The null hypothesis is that the two discordant topologies are equally frequent, consistent with ILS. Rejection of the null suggests introgression.
• Model-Based Inference (Optional): To characterize the introgression event (e.g., direction, timing), input your gene trees or sequence alignments into a phylogenetic network inference tool. These methods will fit a model that includes both ILS and introgression pulses to your data.

Visual Workflows and Diagrams

Phylogenomic Analysis Workflow

ILS vs. Introgression Signals

The Scientist's Toolkit: Research Reagent Solutions

The following table lists key software and data resources essential for conducting phylogenomic analyses focused on introgression.

Item Name	Type	Primary Function
D-statistic (ABBA-BABA)	Statistical Test	A simple, powerful test to detect an asymmetry in gene tree frequencies that is a hallmark of introgression between three ingroup species [1].
Phylogenetic Network Software	Model-based Inference Tool	Software packages (e.g., PhyloNet, SNaQ) used to infer explicit phylogenetic networks that represent evolutionary histories containing both divergence (speciation) and introgression events [1].
ASTRAL	Species Tree Inference Tool	A tool for estimating the primary species tree from a collection of gene trees. It is statistically consistent under the multi-species coalescent model and is useful for establishing a baseline topology against which discordance is measured [60].
Whole Genome Alignment	Data Resource	A genome-wide, base-pair alignment of multiple species. This serves as the fundamental data source from which loci or windows are extracted for gene tree estimation and subsequent introgression analysis [60].

Ensuring Robust Conclusions: Validation Strategies and Cross-Method Comparisons

Frequently Asked Questions (FAQs)

Q1: My D-statistic shows a significant signal of introgression, but my phylogenetic network analysis does not. What could be the cause? A significant D-statistic alone is not always conclusive evidence of introgression. This discrepancy can arise from several factors:

• Confounding Evolutionary Forces: The D-statistic can be confounded by other processes like gene flow from an unsampled "ghost" lineage or regions of the genome under strong selective pressures that mimic introgression signals [1] [52].
• Methodological Differences: The D-statistic detects an excess of shared derived alleles between non-sister taxa, while phylogenetic network methods often model introgression as discrete pulses. If gene flow is continuous or very ancient, network methods may not infer it [1].
• Recommendation: Triangulate with additional methods. Use summary statistics like RNDmin or Gmin that are based on sequence divergence to confirm the presence of recent gene flow in specific genomic regions [48].

Q2: How can I determine if gene tree discordance in my dataset is caused by Incomplete Lineage Sorting (ILS) or introgression? Distinguishing between ILS and introgression is a core challenge. Your analysis should leverage the different genomic signatures of each process:

• Expected Patterns under ILS: Under a model of pure ILS without introgression, the frequencies of the two discordant gene tree topologies are expected to be equal [1]. A significant deviation from this equal frequency, where one discordant topology is more abundant, is a classic signature of introgression.
• Excess Allele Sharing: Methods like the D-statistic formalize this by testing for an excess of shared derived alleles between non-sister taxa, which is not expected under ILS alone [1] [52].
• Sequence Similarity: In introgressed regions, you expect to find haplotypes between species that are highly similar and have a recent coalescent time. Statistics like dmin (the minimum sequence distance between any two haplotypes from different species) can identify these recently shared haplotypes [48].

Q3: What are the best practices for data sharing to ensure my phylogenomic analyses are reproducible? Reproducibility is critical for robust science. Adhere to the following guidelines:

• Share Digital Data, Not Just Figures: Publish your multiple sequence alignments, character matrices, and phylogenetic trees as digital files in a public repository like TreeBASE or Dryad, not just as images in a paper [61].
• Use Consistent Taxon Labels: Ensure the taxon labels in your phylogenetic tree exactly match those in your alignment or character matrix to avoid confusion and enable data integration [61].
• Provide a README File: Include a plain-text README file that describes the contents of your data package, lists all files, and explains any abbreviations or codes used [61].

Troubleshooting Guides

Issue: Low Power to Detect Introgression

Problem: You suspect introgression, but standard tests are not returning significant results.

Potential Causes and Solutions:

• Cause 1: Ancient or Weak Introgression. The signal may have decayed over time.
  • Solution: Use methods sensitive to older gene flow, such as phylogenetic network inference (e.g., using maximum likelihood or Bayesian methods) that can model ancestral introgression events [1] [52].
• Cause 2: Introgression is Rare or Localized. The signal might be diluted if only a small fraction of the genome is introgressed.
  • Solution: Perform window-based or locus-based scans across the genome. Methods like RNDmin or Gmin are designed to identify these specific "islands of introgression" [48].
  • Solution: Ensure your data is phased. Statistics like dmin and Gmin require phased haplotype data to identify the highly similar sequences indicative of recent introgression [48].

Issue: False Positive Signals of Introgression

Problem: Your tests indicate introgression, but you are concerned the signal might be spurious.

Potential Causes and Solutions:

• Cause 1: Variation in Mutation Rate. Genomic regions with a naturally low mutation rate can exhibit high similarity that mimics recent introgression [48].
  • Solution: Use statistics that are robust to mutation rate variation. RNDmin and Gmin normalize for this by comparing within-species divergence to between-species divergence or to an outgroup, effectively controlling for locus-specific mutation rates [48].
• Cause 2: Selection. Positive or background selection can create patterns of high divergence or similarity that confound introgression tests [1] [48].
  • Solution: Triangulate with multiple methods. While some phylogenomic methods are relatively robust to selection, combining topology-based methods (like D-statistics) with divergence-based methods (like RNDmin) can help confirm the signal is due to gene flow [1] [48]. Additionally, check if candidate introgressed regions are enriched for genes under selection.
• Cause 3: Inadequate Taxon Sampling.
  • Solution: Include multiple individuals per species/population and, if possible, sample closely related outgroups. This helps clarify population-level processes and can reveal whether signal is consistent across the population [48].

Key Experimental Protocols

Protocol 1: Conducting a D-Statistic (ABBA-BABA) Test

This test detects introgression by measuring an excess of shared derived alleles between non-sister taxa.

• Sequence Alignment: Generate a whole-genome alignment for your focal taxa (P1, P2, P3) and an outgroup (O).
• Variant Calling: Identify bi-allelic single nucleotide polymorphisms (SNPs) across the alignment.
• Define Phylogeny: Establish the known species relationship as ((P1,P2),P3).
• Site Pattern Counting: Scan the genome and count the frequencies of two site patterns:
  • ABBA: Sites where P2 and P3 share a derived allele, while P1 and O have the ancestral allele.
  • BABA: Sites where P1 and P3 share a derived allele, while P2 and O have the ancestral allele.
• Calculate D-Statistic: Use the formula: D = (Sum(ABBA) - Sum(BABA)) / (Sum(ABBA) + Sum(BABA)).
• Significance Testing: Assess the statistical significance of the D-value using a block jackknife or permutation approach. A significant deviation from zero suggests introgression between P3 and P2 (if D>0) or P3 and P1 (if D<0) [1].

Protocol 2: Identifying Introgressed Loci using RNDmin

This method identifies loci with unusually high similarity between species, indicating recent introgression, while controlling for mutation rate variation.

• Data Preparation: Obtain phased haplotype data for two sister species (X and Y) and an outgroup (O).
• Calculate Divergences: For each genomic window or locus:
  • Compute dmin, the minimum pairwise sequence distance between any haplotype from species X and any haplotype from species Y.
  • Compute dXO, the average sequence distance between species X and the outgroup O.
  • Compute dYO, the average sequence distance between species Y and the outgroup O.
  • Calculate dout = (dXO + dYO)/2.
• Compute RNDmin: Calculate the statistic as RNDmin = dmin / dout [48].
• Identify Outliers: Compare RNDmin values across the genome. Genomic windows with exceptionally low RNDmin values are strong candidates for containing introgressed sequences, as they indicate a much more recent divergence between the closest haplotypes than the average divergence from the outgroup.

Table 1: Comparison of Phylogenomic Methods for Detecting Introgression

Method	Type	Key Principle	Data Required	Strengths	Weaknesses
D-Statistic (ABBA-BABA) [1]	Summary Statistic	Excess of shared derived alleles in a four-taxon quartet.	A single sequence (or allele counts) from each of 3 ingroup taxa + 1 outgroup.	Simple, fast, powerful for detecting recent introgression.	Sensitive to model violations; does not localize introgression in the genome.
Phylogenetic Networks [1] [52]	Model-Based	Co-estimates species phylogeny and introgression events from gene trees.	Genome-wide gene trees or sequence alignments from multiple individuals/species.	Models history directly; can estimate timing and direction of introgression.	Computationally intensive; complex model selection.
RNDmin & Gmin [48]	Summary Statistic	Minimum inter-species sequence distance, normalized by divergence to an outgroup or average distance.	Phased haplotypes from two sister species + an outgroup.	Robust to mutation rate variation; can pinpoint specific introgressed loci.	Requires phased data; most powerful for recent introgression.
Population Branch Statistic (PBS)	Summary Statistic	Measures lineage-specific differentiation, identifying loci with extreme divergence.	Genotype data from multiple individuals from three populations.	Useful for detecting selection and local adaptation following introgression.	Cannot easily distinguish introgression from other causes of low divergence.

Table 2: Key "Research Reagent Solutions" for Introgression Analysis

Item	Function	Considerations
Whole-Genome Sequencing Data	Provides the base pairs for all analyses, allowing for genome-wide scans and high-resolution detection.	Costly for many individuals; computational burden for storage and processing [62].
Reduced-Representation Sequencing (e.g., GBS, RADseq)	Provides a cost-effective way to genotype many individuals across thousands of loci for phylogenetic and population genetic analysis [63].	Captures only a fraction of the genome; loci may not be independent [63].
Phased Haplotype Data	Essential for methods that rely on identifying shared haplotypes between species (e.g., dmin, Gmin) [48].	Requires specialized sequencing protocols or statistical phasing, which can introduce errors.
High-Quality Reference Genome	Enables accurate read mapping, variant calling, and provides genomic context for identified introgressed regions.	Availability may be limited in non-model organisms.
Outgroup Genome Sequence	Crucial for polarizing alleles (as ancestral or derived) in D-statistics and for normalizing divergence in methods like RNDmin [1] [48].	Should be a closely related lineage that diverged before the ingroup.

Analytical Workflows and Visualizations

Diagram: Triangulating Evidence to Distinguish ILS from Introgression

Diagram: Expected Gene Tree Patterns under ILS vs. Introgression

Comparative Analysis of Nuclear, Plastid, and Mitochondrial Genomes

Troubleshooting Guides

Guide 1: Resolving Issues in Organellar Genome Assembly

Problem: Incomplete or Fragmented Mitogenome Assembly

• Symptoms: Assembly results in many short, fragmented contigs rather than a single, circular chromosome or defined multichromosomal structure.
• Primary Causes:
  • High Repetitive Content: Plant mitogenomes are often enriched with long, recombinationally active repeats that short-read technologies cannot resolve [64] [65].
  • Interference from NUMTs/MTPTs: Nuclear sequences of mitochondrial origin (NUMTs) or plastid sequences in the mitogenome (MTPTs) can be mistakenly assembled, leading to misassemblies or loss of true mitochondrial structure [64].
  • Low Sequence Divergence: The extremely low mutation rate of plant mitogenomes can result in insufficient variation for assemblers to resolve repeats [65].
• Solutions:
  • Utilize Long-Read Sequencing: Employ PacBio HiFi or Oxford Nanopore sequencing to generate reads long enough to span repetitive regions [64] [65].
  • Apply Specialized Assemblers: Use assembly tools specifically designed for plant mitogenomes, such as PMAT, which leverages copy number differences to separate mitochondrial from nuclear and plastid reads without pre-filtering [64].
  • Leverage Coverage Differences: Mitochondrial reads often have a different (and typically higher) coverage depth than nuclear reads, which can help in segregating them during assembly [64].

Problem: Phylogenetic Incongruence Between Genomes

• Symptoms: Phylogenetic trees reconstructed from nuclear, plastid, and mitochondrial datasets show conflicting topologies, creating uncertainty in evolutionary relationships.
• Primary Causes:
  • Incomplete Lineage Sorting (ILS): The failure of gene lineages to coalesce in a population deeper than a speciation event, leaving different ancestral polymorphisms in descendant species [64] [65].
  • Introgression/Hybridization: The transfer of genetic material between species through hybridization, leading to a genome history that differs from the species history [65].
  • Different Evolutionary Rates: Mitochondrial genes, especially in plants, evolve much more slowly than plastid or nuclear genes, leading to a lack of phylogenetic signal at shallow timescales [64] [65].
• Solutions:
  • Multi-Species Coalescent Modeling: Use methods like ASTRAL or SVDquartets to account for gene tree discordance when inferring the species tree, which helps distinguish ILS from other causes [64].
  • Test for Discordance: Use statistical tests (e.g., Patterson's D statistic) to formally test for signals of introgression between lineages [65].
  • Leverage Mitogenomes for Deep Phylogeny: For resolving ancient evolutionary relationships, use the slowly evolving mitogenome, as its low substitution rate reduces homoplasy [64].

Guide 2: Troubleshooting Sequencing and Library Preparation

Common issues in Next-Generation Sequencing (NGS) preparation that can impact all genome sequencing projects are summarized in the table below.

Table 1: Troubleshooting Common NGS Library Preparation Issues

Problem Category	Typical Failure Signals	Common Root Causes	Corrective Actions
Sample Input & Quality	Low library complexity, smeared electrophoretogram, low yield	Degraded DNA/RNA; contaminants (phenol, salts); inaccurate quantification [66]	Re-purify input; use fluorometric (Qubit) over UV quantification; check purity ratios (260/230 > 1.8) [66]
Fragmentation & Ligation	Unexpected fragment size; sharp ~70-90 bp peak (adapter dimers)	Over-/under-shearing; improper adapter-to-insert ratio; poor ligase performance [66]	Optimize fragmentation parameters; titrate adapter concentration; ensure fresh enzymes and buffers [66]
Amplification (PCR)	High duplicate rate; amplification bias; overamplification artifacts	Too many PCR cycles; carryover enzyme inhibitors; primer exhaustion [66]	Reduce PCR cycles; use master mixes; optimize annealing conditions [66]
Purification & Cleanup	Sample loss; incomplete removal of adapter dimers; carryover contaminants	Wrong bead-to-sample ratio; over-drying beads; inadequate washing [66]	Precisely follow cleanup protocols; avoid over-drying beads; use fresh wash buffers [66]

Frequently Asked Questions (FAQs)

Q1: Why is the mitochondrial genome so much larger and more complex in plants compared to animals? Plant mitogenomes exhibit an "evolutionary paradox": they have extremely low sequence mutation rates but undergo frequent genomic recombination, often mediated by repetitive sequences [64]. This recombination leads to enormous size variation (from 66 kb to over 18 Mb) and complex structures, including multi-circular chromosomes, linear forms, and branched networks, moving far beyond the simple, compact animal mitogenome model [64] [65].

Q2: How can I determine if phylogenetic incongruence is due to Incomplete Lineage Sorting (ILS) or introgression? Distinguishing between ILS and introgression is a central challenge. ILS is expected to create random and largely symmetrical patterns of discordance across the genome [64] [65]. In contrast, introgression produces a directional signal, where gene trees supporting a specific species relationship are over-represented in genomic regions that were transferred between species. Statistical tests like Patterson's D (ABBA-BABA test) can detect these asymmetrical signals of gene flow, providing evidence for introgression over ILS [65].

Q3: What are the key advantages of using mitochondrial genomes for resolving deep evolutionary relationships? While mitochondrial genes perform poorly for shallow-level phylogenetics due to their slow nucleotide substitution rates (in plants), this feature becomes an advantage for reconstructing ancient lineages [64]. The low mutation rate reduces the risk of homoplasy (reversed or parallel mutations), which can obscure true phylogenetic signal over long evolutionary timescales. Mitogenomes thus provide rich, conserved phylogenetic information that can resolve relationships that are unresolved by faster-evolving plastid or nuclear genes [64].

Q4: What is the difference between NUMTs and MTPTs, and how do they complicate assembly?

• NUMTs (Nuclear Mitochondrial DNA segments) are sequences of mitochondrial origin that have been transferred and integrated into the nuclear genome [65].
• MTPTs (Mitochondrial Plastid DNA segments) are sequences of plastid origin that have been transferred and integrated into the mitochondrial genome [65]. Both can be mistakenly assembled into the organellar genome if using tools that don't properly account for them, leading to misassembled chimeric sequences. Specialized assemblers that use coverage depth and k-mer frequency can help separate these sequences [64].

Experimental Protocols

Protocol 1: Assembling a Complex Plant Mitogenome using PMAT

This protocol is adapted from recent studies on assembling multichromosomal mitogenomes [64].

1. DNA Extraction and Sequencing

• Material: Fresh plant tissue (e.g., leaves).
• Extraction: Use a high-quality DNA extraction kit (e.g., Hi-DNAsecure Plant Kit) to obtain high-molecular-weight, high-integrity genomic DNA. Verify quality via agarose gel electrophoresis and spectrophotometry (e.g., Nanodrop).
• Sequencing: Prepare a 15-kb sequencing library (e.g., using SMRTbell Express Template Prep Kit 2.0) and sequence on a long-read platform (e.g., PacBio Revio) to generate HiFi reads [64].

2. Mitogenome Assembly with PMAT

• Input: PacBio HiFi sequencing data in FASTA or FASTQ format.
• Software: PMAT2 (Plant Mitochondrial Assembly Tool).
• Command:
  • -t hifi: Specifies the input data type as HiFi reads.
  • -T 50: Uses 50 CPU threads for faster computation [64].
• Output: The primary output is an assembly graph in GFA format.

3. Graph Visualization and Disentanglement

• Software: Bandage.
• Procedure:
  • Load the GFA file into Bandage.
  • Visually inspect the graph for circular structures and potential branching.
  • Use Bandage to disentangle the graph and extract complete circular chromosomes [64].

4. Genome Annotation

• Initial Annotation: Use the online platform PMGA for initial gene prediction.
• tRNA Validation: Run tRNAscan-SE v2.0 to identify and validate tRNA genes.
• Manual Curation: Use software like MacVector to manually inspect and correct all annotated genes (PCGs, rRNAs, tRNAs) for accuracy [64].

Protocol 2: A Workflow for Testing ILS vs. Introgression

1. Data Collection and Orthology Assignment

• Sequence and assemble data from all three genomes (nuclear, plastid, mitochondrial) for multiple individuals across the species group of interest.
• Identify orthologous genes or single-copy nuclear genes using tools like OrthoFinder.

2. Individual Gene Tree Inference

• For each orthologous gene or genomic region, infer a phylogenetic tree using maximum likelihood (e.g., with IQ-TREE) or Bayesian methods (e.g., with MrBayes).

3. Species Tree Inference and Discordance Analysis

• Primary Species Tree: Infer a species tree using coalescent-based methods (e.g., ASTRAL-III) that account for ILS by modeling individual gene trees as inputs.
• Visualize Discordance: Use a software like DiscoVista to visualize the distribution of different topological relationships among gene trees.

4. Test for Introgression

• D-Statistic (ABBA-BABA Test): Use a tool like Dsuite to calculate Patterson's D statistic. This test evaluates an excess of shared derived alleles between two species to a third, which is a signature of introgression [65].
• Phasing Data: For the most robust results, use phased haplotype data for these tests.

5. Synthesis

• Interpret the results collectively. Widespread, random discordance is consistent with ILS, while significant, directional signals from D-statistics support introgression.

Workflow and Relationship Visualizations

Figure 1: Decision workflow for distinguishing ILS from Introgression.

Figure 2: Organellar genome assembly workflow.

The Scientist's Toolkit

Table 2: Essential Research Reagents and Tools for Comparative Organellar Genomics

Tool / Reagent	Function / Description	Application in Comparative Genomics
PacBio Revio / Sequel	Long-read sequencer generating HiFi reads.	Essential for resolving complex, repeat-rich regions in mitogenomes and large structural variants [64] [65].
PMAT (v2)	A specialized assembler for plant mitogenomes.	Uses copy number differences to separate organellar from nuclear reads, avoiding NUMT/MTPT interference [64].
Bandage	A graphical tool for visualizing assembly graphs.	Allows for manual inspection and disentanglement of complex mitogenome structures from assembly graphs [64].
ASTRAL-III	Software for species tree inference from gene trees under the multi-species coalescent model.	Infers the primary species tree while accounting for discordance caused by ILS [64].
Dsuite	A software package for calculating D-statistics.	Used to test for signals of introgression between species by analyzing allele frequency patterns [65].
Hi-DNAsecure Plant Kit	A DNA extraction kit designed for high-quality, high-molecular-weight DNA.	Provides the input material necessary for successful long-read sequencing of all genomic compartments [64].

Troubleshooting Guide: FAQs on Distinguishing ILS from Introgression

Q1: My genomic data shows significant gene tree discordance. How can I determine if the cause is Incomplete Lineage Sorting (ILS) or introgression? A1: To distinguish between these processes, analyze the frequencies of different gene tree topologies across your genome-wide data [1]. Under a pure ILS scenario, the two discordant topologies are expected to be equal in frequency. An significant excess of one discordant topology is a key signature of introgression [1]. You can use methods like the D-statistic (ABBA-BABA test) to test for this imbalance statistically.

Q2: I am studying a hybrid zone between two pine species. What is the expected genomic pattern of adaptive introgression? A2: Adaptive introgression appears as genomic regions where gene flow from one species into another is higher than the genome-wide background level, and these regions are associated with adaptive traits or local environmental conditions [67]. In pines, this often involves loci related to stress tolerance (e.g., for bog habitats) that introgress from the pre-adapted species into the other [67].

Q3: What are the minimum data requirements for a reliable test of introgression? A3: The minimum requirement is genomic data (e.g., from whole-genome sequencing) from a single haploid individual each from three focal species or populations, plus an outgroup [1]. This "quartet" or "rooted triplet" forms the basis for many phylogenomic tests, including those based on the multispecies coalescent model [1].

Q4: How can ecological data be integrated with genomic analyses to validate adaptive introgression? A4: Ecological Niche Modeling (ENM) can be used to project species' distributions under past or present climates. When combined with genomic scans for selection, you can test whether introgressed alleles are associated with specific environmental variables (e.g., moisture, temperature) that define the ecological niche of the donor species. Finding that introgressed regions confer adaptation to the recipient species' niche, particularly in marginal habitats, provides strong validation [67].

Q5: Could selection alone mimic the signals of introgression? A5: While both selection and demography can create similar patterns locally, most phylogenomic methods that use genealogical signals (like gene tree frequencies and branch lengths) are robust to the effects of selection [1]. However, it is always recommended to use multiple complementary methods and to validate findings with ecological and functional data [1].

Quantitative Data from Pine Systematics Case Study

The following tables summarize key quantitative findings from genomic studies on Pinus sylvestris and P. mugo hybrid zones [67].

Table 1: Sampling and Hybrid Classification in Pine Contact Zones

Location	Population Code	Habitat Type	Sample Size	Putative Pure Species	F1 Hybrids	Advanced Backcrosses
Bór na Czerwonem	BC	Peat Bog	Not Specified	P. sylvestris, P. mugo	Present	Majority shifted towards P. mugo ancestry
Błędne Skały	BS	Sandstone Formations	Not Specified	P. sylvestris, P. mugo	Present	Majority shifted towards P. mugo ancestry
Torfowisko pod Zieleńcem	TZ	Peat Bog	Not Specified	P. sylvestris, P. mugo	Present	Majority shifted towards P. mugo ancestry
Reference Stands	N/A	Allopatric	1,558 total (24 pops)	12 PS, 9 PM	Absent	Absent

Table 2: Genomic Signals of Selection and Introgression

Analysis Type	Key Finding	Implication
Outlier Loci (Selection Scan)	Most outlier loci were shared across all sympatric populations, but some were unique to individual contact zones.	Indicates a combination of globally and locally important adaptive pressures.
Biological Function of Outliers	Mainly associated with regulatory processes: phosphorylation, proteolysis, and transmembrane transport.	Introgression affects key adaptive physiological and signaling pathways.
Strength of Selection Signal	Strongest in pure P. sylvestris and hybrids with majority P. sylvestris ancestry. Weaker in P. mugo individuals.	Suggests P. mugo was pre-adapted to peat bog habitats, and introgression aids P. sylvestris adaptation at its ecological margin.

Experimental Protocols for Phylogenomic Analysis

Protocol 1: Population Sampling and SNP Genotyping for Hybrid Zone Analysis

Objective: To characterize patterns of interspecific gene flow and identify genomic regions under selection.

Materials:

• Plant material (needles or cambium) from allopatric parental populations and sympatric contact zones.
• Permits for field collection (e.g., from relevant environmental ministries).
• DNA extraction kit suitable for plant tissue.
• SNP genotyping platform (e.g., microarray, targeted sequence capture) for thousands of nuclear markers.

Methodology:

• Sampling: Collect tissue samples from a large number of individuals (e.g., >1500) from reference allopatric populations of both parent species and from multiple hybrid zones [67].
• DNA Extraction: Perform high-quality DNA extraction. The specific method will depend on the plant tissue and downstream genotyping application.
• Genotyping: Use a high-throughput method to genotype all individuals at thousands of nuclear Single Nucleotide Polymorphisms (SNPs) [67].
• Data Quality Control: Filter SNPs and individuals based on missing data and minor allele frequency.

Protocol 2: Distinguishing ILS from Introgression Using Quartet Sampling

Objective: To test for significant introgression between a pair of sister species using a third species and an outgroup.

Materials:

• Whole-genome sequencing data or genome-wide SNP data from a single individual (haploid is sufficient) from each of three ingroup species (P1, P2, P3) and one outgroup (O).
• Computational resources and software for phylogenetic analysis (e.g., D-statistic calculation).

Methodology:

• Gene Tree Estimation: Estimate gene trees from alignments of many independent loci or non-overlapping genomic windows across the genome [1].
• Count Topologies: For each locus, record which of the three possible unrooted quartet topologies it supports: Topo1: ((P1,P2),P3,O), Topo2: ((P1,P3),P2,O), or Topo3: ((P2,P3),P1,O) [1].
• Apply D-Statistic: Use the D-statistic to test for an imbalance in the frequency of the two discordant topologies. The statistic is formulated as D = (NTopo2 - NTopo3) / (NTopo2 + NTopo3). A significant deviation from zero indicates introgression [1].
• Interpretation: A significant positive D suggests introgression between P2 and P3, while a significant negative D suggests introgression between P1 and P3.

Workflow Visualization

Workflow for Integrating Genomic and Ecological Analyses

Gene Flow Pattern in a Three-Taxon System

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Materials for Phylogenomic Studies in Non-Model Organisms

Item / Reagent	Function / Application	Context from Case Study
High-Quality DNA Extraction Kit	To obtain pure, high-molecular-weight DNA from tough plant tissues (e.g., pine needles) for downstream genotyping.	Essential for preparing samples for SNP genotyping from over 1,500 individual trees [67].
SNP Genotyping Array / Targeted Capture Panel	To generate genome-wide data on thousands of Single Nucleotide Polymorphisms (SNPs) without needing a full genome sequence for every individual.	The core technology used to genotype individuals at "thousands of nuclear SNPs" [67].
Computational Software (e.g., for D-statistic)	To perform statistical tests for introgression and calculate other population genetic parameters from large SNP datasets.	Required for analyzing gene tree frequencies and testing the null hypothesis of ILS [1].
Reference Genomes (if available)	To align sequencing reads, call variants, and perform functional annotation of genomic regions identified as under selection.	While not explicitly mentioned, a reference is invaluable for annotating the biological functions of outlier loci associated with, e.g., phosphorylation [67].
Ecological Niche Modeling Software (e.g., MaxEnt)	To model species' distributions based on occurrence data and environmental layers, helping to link genotypes to environments.	Used to provide the ecological context for validating adaptive introgression in marginal habitats like peat bogs [67].

A core challenge in modern bacterial phylogenomics is distinguishing between different sources of evolutionary signal. Two primary causes of gene tree discordance are:

• Incomplete Lineage Sorting (ILS): The failure of ancestral genetic polymorphisms to coalesce (merge into a common ancestor) in the immediate ancestral population of species. This leads to gene trees that differ from the species tree by chance, not interbreeding [1].
• Introgression: The transfer of genetic material between distinct species through homologous recombination, blurring species boundaries. In bacteria, this refers to gene flow between the core genomes of different species, analogous to processes in sexual organisms [18].

The following sections provide a technical framework for researchers aiming to design experiments, troubleshoot analyses, and correctly interpret results in this complex field.

Frequently Asked Questions & Troubleshooting Guides

Conceptual and Experimental Design FAQs

Q1: What is the fundamental difference between ILS and introgression as causes of phylogenetic discordance? Both processes create incongruence between gene trees and the species tree, but their underlying mechanisms differ. ILS is a stochastic process arising from the retention of ancestral genetic variation during rapid speciations, producing discordant trees with relatively equal frequencies. In contrast, introgression results from direct gene flow between species, often generating asymmetric patterns where one discordant tree topology is significantly overrepresented [1] [7].

Q2: My phylogenomic analysis shows widespread gene tree discordance. What is the first step in determining if ILS or introgression is the cause? Begin by calculating site concordance factors (sCF) and discordance factors (sDF1/sDF2). High or imbalanced sDF1/sDF2 values at a particular phylogenetic node are a strong indicator to proceed with more specific tests for introgression, such as D-statistics and phylogenetic network analyses [7].

Q3: Are there specific bacterial lineages where introgression is more common? Yes, the prevalence of introgression varies. Studies across 50 major bacterial lineages show that genera like Escherichia–Shigella and Cronobacter exhibit the highest levels of introgression, while other lineages remain more phylogenetically distinct. The process is most frequent between closely related species [18].

Troubleshooting Data Analysis

Q4: My D-statistics are significant, suggesting introgression. Could something else be causing this signal? While a significant D-statistic (e.g., |D| > 0) can indicate introgression, it is crucial to rule out other factors. Significant signals can sometimes be caused by reference bias or mapping errors, especially in analyses involving ancient DNA or low-coverage genomes. Always validate findings with complementary methods like QuIBL, which can help infer the timing and mode of introgression, or phylogenetic network models [1] [7].

Q5: I suspect my bacterial transformation for a functional genomics assay has failed because I see no colonies. What should I check?

• Competent Cells: Verify cell viability and transformation efficiency by plating on non-selective media and including a positive control with a known, high-quality plasmid [68] [69].
• DNA Quality and Quantity: Ensure you are using the recommended amount of pure, contaminant-free DNA (e.g., 1-10 ng for 50-100 µL of chemically competent cells). Avoid excessive amounts of DNA [68].
• Selection Conditions: Confirm that your selective plates contain the correct antibiotic corresponding to your vector's resistance marker and that the antibiotic has not degraded [68] [69].

Q6: When using IQ-TREE for phylogenomic analysis, what support values should I trust for single genes versus concatenated datasets?

• For single gene trees, you can generally rely on a branch with UFBoot ≥ 95% and SH-aLRT ≥ 80% [70].
• For concatenated phylogenomic datasets, standard bootstrap supports (including UFBoot) can be overly optimistic and tend toward 100%. In these cases, it is highly recommended to compute concordance factors to get a more realistic measure of branch support given the underlying gene tree heterogeneity [70].

Quantitative Data on Bacterial Gene Flow and Introgression

The tables below summarize key quantitative findings from recent large-scale genomic studies, providing benchmarks for your own research.

Table 1: Prevalence of Clonality and Recombination in Bacteria

Evolutionary Pattern	Prevalence (%)	Notes and Examples
Truly Clonal Species	2.6% - 12.8%	2.6% identified by two methods; 12.8% by at least one method. Often endosymbionts (e.g., Chlamydia, Brucella) [71].
Species with Recombination	~90%	The vast majority of bacterial species show clear signs of homologous recombination [71].
Homoplasies in Core Genome	Average of 35 per core gene	Pervasive across core genomes; most are synonymous, suggesting recombination, not selection, is the primary driver [71].

Table 2: Measured Levels of Introgression Across Bacterial Lineages

Metric	Value	Context
Average Introgressed Core Genes	2% (Median)	Across 50 bacterial genera [18].
Maximum Reported Introgression	Up to 14%	Found in the Escherichia–Shigella lineage [18].
High Introgression Example	~33%	Initially observed between two Streptococcus parasanguinis ANI-defined species, but they were later reclassified as a single Biological Species Concept (BSC) species [18].
Divergence Barrier for Gene Flow	Generally 90-98% genome identity	The ~95% ANI species definition is an approximation of where gene flow is often interrupted [71].

Detailed Experimental Protocols

Protocol 1: A Basic Phylogenomic Pipeline to Detect Introgression

This protocol outlines a standard workflow for detecting introgression from genomic data [7] [18].

• Genome Sequencing and Data Collection: Assemble a dataset of whole-genome sequences for your target taxa, including multiple genomes per putative species and appropriate outgroups.
• Species Delimitation: Define initial species boundaries using an Average Nucleotide Identity (ANI) threshold of 94-96% for core genomes. This creates operational "ANI-species" [18].
• Core Genome Alignment: Identify and extract single-copy orthologous genes present in all (or most) genomes. Create a multiple sequence alignment for each gene and/or a concatenated alignment of all core genes.
• Phylogenetic Inference:
  • Infer a maximum-likelihood (ML) tree from the concatenated core genome alignment. This serves as the reference species tree.
  • Infer individual ML gene trees from each of the single-gene alignments.
• Detect Gene Tree Discordance: Use the reference species tree and the set of gene trees to calculate site concordance factors (sCF) and discordance factors (sDF). Visually inspect nodes with high or imbalanced sDF values.
• Test for Introgression:
  • Apply the D-statistic (ABBA-BABA test) to rooted quartets of taxa to test for significant asymmetry in site patterns indicative of introgression.
  • Use tools like QuIBL to infer the timing and mode of putative introgression events.
  • Perform phylogenetic network analysis (e.g., with PhyloNet or SplitsTree) to model potential reticulate evolutionary events.
• Refine Species Boundaries (Optional): Re-analyze patterns of gene flow using methods like homoplasic-to-non-homoplasic allele (h/m) ratios to refine ANI-species into more biologically grounded "BSC-species," which may reduce perceived introgression between very closely related groups [71] [18].

Protocol 2: Differentiating ILS from Introgression Using Site Frequencies

This protocol leverages the expected frequencies of different gene tree topologies under a model of pure ILS [1].

• Select a Rooted Quartet: Choose four taxa for analysis: three ingroup taxa (P1, P2, P3) whose relationships are of interest, and one outgroup (O) for rooting.
• Count Gene Tree Topologies: For a large set of genomic loci (e.g., genes or windows), tally the frequency of the three possible unrooted topologies relating P1, P2, and P3.
• Compare to Null Expectation: Under a scenario of pure ILS without introgression, the two discordant topologies (those where P2 and P3 are closest, or P1 and P3 are closest) are expected to occur at equal frequencies.
• Interpret Results: A significant deviation from this equal frequency, with one discordant topology being much more common than the other, is evidence for introgression between the specific taxa united by the overrepresented topology.

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Reagents and Software for Phylogenomic Analysis

Item Name	Function/Application	Usage Notes
Competent Cells (e.g., commercial E. coli strains)	Propagating plasmid DNA for sequencing or functional assays.	Check genotype (e.g., recA mutation to prevent recombination), transformation efficiency, and antibiotic resistance. Store at -70°C, avoid freeze-thaw cycles [68].
Selection Antibiotics (e.g., Ampicillin, Kanamycin)	Selecting for transformants containing your plasmid vector.	Verify correct antibiotic corresponds to vector resistance. Use carbenicillin instead of unstable ampicillin for more stable selection [68].
IQ-TREE Software	Phylogenetic inference using maximum likelihood.	Performs model selection, tree inference, ultrafast bootstrap (UFBoot), and calculation of concordance factors. Use -bb for UFBoot and -alrt for SH-aLRT test [70].
ASTRAL Software	Inferring species trees from multiple gene trees under the multi-species coalescent (MSC) model.	Accounts for ILS; useful for constructing a reference species tree in the presence of gene tree discordance [7].
D-Statistic	Detecting allele-sharing asymmetry indicative of introgression in a four-taxon setup.	A significant	D	value suggests introgression. Implemented in tools like HYBRIDCHECK or as part of larger phylogenomic packages.
Phylogenetic Networks (e.g., PhyloNet, SplitsTree)	Visualizing and modeling evolutionary relationships that are not strictly tree-like, including hybridization and introgression.	Used to test whether a network model fits the data significantly better than a strictly bifurcating tree model [7].

Workflow Visualization for Distinguishing ILS and Introgression

The following diagram illustrates a logical workflow for investigating the source of gene tree discordance, incorporating the methods discussed above.

Diagram 1: A logical framework for distinguishing ILS from introgression.

Frequently Asked Questions

1. What are the primary biological processes that cause gene tree discordance? Gene tree discordance, where gene trees conflict with the species tree, is primarily caused by three biological processes: Incomplete Lineage Sorting (ILS), introgression (hybridization), and independent mutations. ILS and introgression are considered the most frequent drivers of this discordance, especially in rapidly radiating groups [72].

2. How can I determine if observed phylogenetic discordance is due to ILS or introgression? Distinguishing between ILS and introgression requires a multi-faceted approach. ILS is more common when speciation events occur in rapid succession and effective population sizes are large, leading to the retention of ancestral genetic variation. Introgression involves the exchange of genetic material after speciation. Methods like D-statistics and phylogenetic networks can identify footprints of introgression, while the pattern of gene tree heterogeneity can provide clues about ILS [72] [73].

3. My analysis shows high gene tree heterogeneity. What does this mean? High gene tree heterogeneity indicates that a significant proportion of genes in your dataset tell a different evolutionary story from the species tree. This is a common feature in groups that have undergone rapid evolutionary radiations. This heterogeneity can be driven by both ILS and introgression, and specialized methods are needed to dissect their relative contributions [72] [73].

4. What is the "anomaly zone" and why is it important? The anomaly zone is a theoretical region of a species tree where, due to very short internal branches (rapid speciation), the most frequently occurring gene tree topology is statistically inconsistent with the species tree topology. In such cases, even the most common gene tree is the "wrong" one. This is crucial because it means that simple concatenation of genomic data can produce a strongly supported but incorrect species tree [73].

5. When should I use a multispecies coalescent network (MSCN) model instead of a simple species tree? You should consider using an MSCN approach when there is strong evidence of gene flow between lineages, such as significant results from D-statistics or other tests for introgression. MSCN models jointly account for both ILS and introgression (reticulation), providing a more biologically realistic framework for groups with a history of hybridization [73].

6. What are the best practices for benchmarking different phylogenetic methods? Best practices include:

• Using Multiple Method Categories: Apply methods from different categories (summary statistics, probabilistic modeling, supervised learning) to the same dataset to cross-validate results [52].
• Systematic Benchmarking: Compare the performance of methods on simulated datasets where the true evolutionary history is known. This helps assess accuracy and power [52].
• Transparent Analysis: Clearly report the methods, software versions, and parameters used to ensure reproducibility [52].

Experimental Protocols for Key Analyses

Protocol 1: Conducting D-Statistic (ABBA-BABA) Tests for Introgression

Purpose: To test for evidence of gene flow between a set of four taxa (((P1,P2),P3),Outgroup).

Methodology:

• Taxon Selection: Define your four populations or species: P1, P2, P3, and an Outgroup. The test assesses if P3 shares more derived alleles with P2 than with P1, which is a signal of introgression between P3 and P2.
• Variant Calling: Use whole-genome resequencing or SNP array data to generate genotype calls for all four taxa.
• Site Identification: Scan the genome to identify sites that are polymorphic across the four taxa. Classify sites into ABBA (where P2 and P3 share a derived allele not found in P1) and BABA (where P1 and P3 share a derived allele not found in P2) patterns.
• Calculation: Compute the D-statistic using the formula: D = (∑ABBA - ∑BABA) / (∑ABBA + ∑BABA). A significant deviation from zero (assessed via block jackknifing) indicates introgression.
• Software: This test is implemented in software packages like Dsuite or ADMIXTOOLS.

Protocol 2: Inferring a Species Tree under the Multispecies Coalescent

Purpose: To estimate the dominant evolutionary history of a group of species while accounting for gene tree heterogeneity caused by ILS.

Methodology:

• Locus Selection: Generate a set of multiple, unlinked, and neutrally evolving sequence alignments (e.g., from exons, UCEs, or whole-genome scaffolds).
• Gene Tree Estimation: Infer a maximum likelihood or Bayesian phylogenetic tree for each individual locus.
• Species Tree Inference: Use a coalescent-based method to estimate the species tree from the distribution of gene trees. These methods model the coalescent process within each branch of the species tree.
• Software: Popular implementations include ASTRAL (a summary method that uses gene trees as input) and SNAPP (a full-likelihood Bayesian method that uses biallelic markers as input).

Protocol 3: Constructing Phylogenetic Networks with MSCN

Purpose: To infer an evolutionary history that includes both divergence (tree-like) and hybridization/introgression (reticulation) events.

Methodology:

• Input Data: Use genome-wide SNP or sequence data, similar to species tree inference.
• Model Selection: Choose a multispecies coalescent network (MSCN) model that incorporates both ILS and introgression. The model will search for a network, rather than a tree, that best explains the genomic data.
• Inference: Run a Bayesian or maximum likelihood analysis to estimate the network topology, branch lengths, and hybridization parameters (e.g., inheritance probabilities, γ).
• Software: Tools like PhyloNet or BEAST with specific packages can be used to infer phylogenetic networks.

Summarizing Key Methodological Approaches

Table 1: Categories of Methods for Analyzing Introgression and ILS

Method Category	Key Principle	Strengths	Common Tools / Metrics
Summary Statistics	Computes patterns of allele sharing or divergence from genomic data.	Fast, easy to compute and interpret, good for initial screening.	D-statistics, f4-ratio, ƒd
Probabilistic Modeling	Uses explicit models of evolutionary processes (coalescent, mutation) to compute the probability of the data.	Provides a powerful statistical framework, can yield fine-scale insights, jointly estimates multiple parameters.	ASTRAL, SNAPP, PhyloNet, BPP
Supervised Learning	Trains algorithms on datasets where the evolutionary history is known to detect patterns in new data.	Emerging approach with great potential, can handle complex patterns when framed as a classification task.	Methods treating introgression detection as a semantic segmentation task [52]

Table 2: Interpreting Signals of Gene Tree Discordance

Observation	Potential Cause	Recommended Follow-up Analysis
Widespread gene tree heterogeneity with short internal branches on the species tree.	Incomplete Lineage Sorting (ILS) driven by rapid speciation.	Use a multispecies coalescent model (e.g., ASTRAL); test for the anomaly zone.
Gene tree heterogeneity that is localized to specific genomic regions or taxa.	Introgression between non-sister lineages.	Perform D-statistics and related tests; use phylogenetic network inference (e.g., PhyloNet).
Gene tree heterogeneity with a strong phylogenetic signal for a specific alternative topology.	Introgression or presence in the anomaly zone.	Analyze gene tree frequencies for asymmetry (suggests introgression); simulate data under different scenarios.
Long branches without extensive ILS between clades with similar phenotypes.	Independent (novel) mutations as a source of convergent evolution [72].	Conduct genome scans for selection; perform functional genetic studies.

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Reagents and Computational Tools for Phylogenomic Analysis

Item / Resource	Function / Purpose
Whole-Genome Resequencing Data	Provides the dense genome-wide marker set required to detect patterns of ILS and introgression.
Reference Genome Assembly	Serves as a scaffold for aligning sequencing reads and calling variants.
Variant Call Format (VCF) File	A standard file format storing genotype information for all samples across genomic positions.
High-Performance Computing (HPC) Cluster	Essential for running computationally intensive phylogenomic analyses on large datasets.
Multispecies Coalescent Model Software (e.g., ASTRAL)	Infers the species tree from a set of gene trees while accounting for ILS.
Phylogenetic Network Software (e.g., PhyloNet)	Infers evolutionary histories that include hybridization events.
Population Genetic Analysis Packages (e.g., ADMIXTOOLS, Dsuite)	Performs tests for admixture and introgression, such as the D-statistic.

Visualizing Workflows and Logical Relationships

Conclusion

Distinguishing between ILS and introgression requires a multifaceted approach that combines robust statistical methods, appropriate model selection, and careful biological interpretation. The integration of phylogenetic networks with traditional tree-thinking, along with validation through comparative genomics and independent data sources, provides the most reliable path forward. For biomedical research, these distinctions are not merely academic—accurate evolutionary histories enable precise identification of conserved drug targets in pathogens, understanding of antibiotic resistance gene flow, and informed conservation strategies for medicinal species. Future directions will likely involve improved model integration, machine learning applications to handle genomic-scale data, and the development of unified reporting standards that facilitate meta-analyses across diverse taxa, ultimately strengthening the evolutionary foundation upon which drug discovery and development rely.

References

• 1. Phylogenomic approaches to detecting and characterizing ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9208645/]
• 2. Teasing apart the sources of phylogenetic tree discordance ... [https://bmcplantbiol.biomedcentral.com/articles/10.1186/s12870-025-06963-3]
• 3. Incomplete lineage sorting and introgression among genera and... [https://pubmed.ncbi.nlm.nih.gov/40296048/]
• 4. Impact of incomplete lineage sorting and natural selection on the... [https://as-botanicalstudies.springeropen.com/articles/10.1186/s40529-025-00482-y]
• 5. Disentangling Sources of Gene Tree Discordance in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7875436/]
• 6. SINE-Based Phylogenomics Reveal Extensive ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8951037/]
• 7. Incomplete lineage sorting and introgression among genera ... [https://bmcbiol.biomedcentral.com/articles/10.1186/s12915-025-02204-z]
• 8. Mechanisms, detection, and impact of horizontal gene transfer ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12448838/]
• 9. Differential gene retention as an evolutionary mechanism ... [https://www.nature.com/articles/srep11571]
• 10. Horizontal Gene Transfer and the History of Life - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC4817804/]
• 11. Introgression - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC10581619/]
• 12. Introgression [https://en.wikipedia.org/wiki/Introgression]
• 13. The contribution of incomplete lineage sorting and ... [https://www.sciencedirect.com/science/article/abs/pii/S1055790322002068]
• 14. Reticulate evolutionary history and extensive introgression ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5146328/]
• 15. Towards Reliable Detection of Introgression in the Presence ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11639170/]
• 16. Chapter 5 Phylogenetic Tree Annotation [https://yulab-smu.top/treedata-book/chapter5.html]
• 17. Add color information to nodes in a phylogenetic tree [https://github.com/acorg/phylo-color]
• 18. Introgression impacts the evolution of bacteria, but species ... [https://www.nature.com/articles/s41467-025-64947-1]
• 19. ABBA BABA statistics [https://evomics.org/learning/population-and-speciation-genomics/2018-population-and-speciation-genomics/abba-baba-statistics/]
• 20. D-statistics for Dummies: A simple test for introgression [https://avianhybrids.wordpress.com/2019/11/09/d-statistics-for-dummies-a-simple-test-for-introgression/]
• 21. Dsuite - Fast D-statistics and related admixture evidence ... [https://pubmed.ncbi.nlm.nih.gov/33012121/]
• 22. millanek/Dsuite: Fast calculation of Patterson's D (ABBA- ... [https://github.com/millanek/Dsuite]
• 23. ipyrad-analysis toolkit: abba-baba - Read the Docs [https://ipyrad.readthedocs.io/en/latest/API-analysis/cookbook-abba-baba.html]
• 24. Abbababa - angsd [https://www.popgen.dk/angsd/index.php/Abbababa]
• 25. Gene flow analysis method, the D-statistic, is robust in a wide ... [https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-017-2002-4]
• 26. The contribution of incomplete lineage sorting and ... [https://pubmed.ncbi.nlm.nih.gov/35905819/]
• 27. Widespread incomplete lineage sorting and introgression ... [https://www.sciencedirect.com/science/article/pii/S2590346223002742]
• 28. Distinguishing Phylogenetic Level-2 Networks with... - DeepPaper [https://arxiv.deeppaper.ai/papers/2507.17308v1]
• 29. Distinguishing phylogenetic level-2 networks with quartets and... [https://ueaeprints.uea.ac.uk/id/eprint/100829/]
• 30. Importance of incomplete lineage sorting and introgression ... [https://www.nature.com/articles/hdy201672]
• 31. Common Methods for Phylogenetic Tree Construction and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11117635/]
• 32. Comparing inference under the multispecies coalescent ... [https://www.sciencedirect.com/science/article/abs/pii/S1055790323000246]
• 33. Detecting Introgression in Shallow Phylogenies [https://pmc.ncbi.nlm.nih.gov/articles/PMC12485364/]
• 34. Implementing and testing the multispecies coalescent model [https://www.sciencedirect.com/science/article/pii/S1055790315003309]
• 35. A Practical Guide to Design and Assess a Phylogenomic Study [https://pmc.ncbi.nlm.nih.gov/articles/PMC9452790/]
• 36. Analysis of phylogenomic datasets reveals conflict ... [https://bmcecolevol.biomedcentral.com/articles/10.1186/s12862-015-0423-0]
• 37. Quantification of congruence among gene trees with ... [https://pubmed.ncbi.nlm.nih.gov/37096985/]
• 38. Phylogenomic approaches untangle early divergences and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9040247/]
• 39. Genomic architecture and introgression shape a butterfly ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7197882/]
• 40. Phylogenomic approach reveals strong signatures of ... [https://www.sciencedirect.com/science/article/pii/S1055790321001330]
• 41. On the inference of complex phylogenetic networks by Markov ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8445492/]
• 42. Subfamily evolution analysis using nuclear and chloroplast ... [https://www.nature.com/articles/s41598-024-83292-9]
• 43. Identifying and addressing methodological incongruence ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10286231/]
• 44. Model Choice, Missing Data and Taxon Sampling Impact ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7115942/]
• 45. The Roles of Protein Structure, Taxon Sampling, and ... [https://www.mdpi.com/2673-4125/1/2/8]
• 46. Detection of Ghost Introgression Requires Exploiting ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11129598/]
• 47. A Guide to Carrying Out a Phylogenomic Target Sequence ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7047930/]
• 48. Powerful methods for detecting introgressed regions from ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4899106/]
• 49. Beyond Level-1: Identifiability of a Class of Galled Tree-Child ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12546270/]
• 50. Impact of incomplete lineage sorting and natural selection on ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12528535/]
• 51. Standing genetic variation and introgression shape the ... [https://www.nature.com/articles/s42003-025-08120-w]
• 52. Review Decoding genomic landscapes of introgression [https://www.sciencedirect.com/science/article/pii/S0168952525001660]
• 53. Best Color Palettes for Scientific Figures and Data ... [https://www.simplifiedsciencepublishing.com/resources/best-color-palettes-for-scientific-figures-and-data-visualizations]
• 54. Color Contrast Accessibility: Complete WCAG 2025 Guide ... [https://www.allaccessible.org/blog/color-contrast-accessibility-wcag-guide-2025]
• 55. 物种树构建后的基因渗入检验方法 [https://cn.bio-protocol.org/mv2/e1010678]
• 56. 国科大博士生导师雷富民研究组揭示基因流在物种形成过程 ... [https://news.ucas.ac.cn/kydt/a167cfc0e1464dcca63d0ecb198ec759.htm]
• 57. 8篇Science齐发| 生命演化研究中心张国捷团队联合中外 ... [https://evolution.zju.edu.cn/science/detail45.html]
• 58. 多物种溯祖模型与物种树推断 - MSE Lab [http://www.mselab.cn/detail/104/]
• 59. Understanding Success Criterion 1.4.11: Non-text Contrast [https://www.w3.org/WAI/WCAG21/Understanding/non-text-contrast.html]
• 60. Primate phylogenomics uncovers multiple rapid... | PLOS Biology [https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3000954]
• 61. Best Practices for Data Sharing in Phylogenetic Research [http://currents.plos.org/treeoflife/article/best-practices-for-data-sharing-in-phylogenetic-research/]
• 62. Genomic evidence for human-mediated introgressive ... [https://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-024-10259-5]
• 63. Data‐driven guidelines for phylogenomic analyses using ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11610416/]
• 64. Assembly and comparative analysis of the mitochondrial ... [https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2025.1704888/full]
• 65. The cucumber (Cucumis sativus L.) mitochondrial genome [https://pmc.ncbi.nlm.nih.gov/articles/PMC12542108/]
• 66. How to Troubleshoot Sequencing Preparation Errors (NGS ... [https://www.cd-genomics.com/resource-sequencing-preparation-troubleshooting.html]
• 67. Molecular signatures of adaptive introgression and selection ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12542055/]
• 68. Bacterial Transformation Troubleshooting Guide [https://www.thermofisher.com/us/en/home/life-science/cloning/cloning-learning-center/invitrogen-school-of-molecular-biology/molecular-cloning/transformation/bacterial-transformation-troubleshooting-guide.html]
• 69. Bacterial Transformation Troubleshooting for Beginners [https://bitesizebio.com/37681/bacterial-transformation-troubleshooting/]
• 70. Frequently Asked Questions - IQ-TREE [https://iqtree.github.io/doc/Frequently-Asked-Questions]
• 71. Gene flow and introgression are pervasive forces shaping the evolution of bacterial species | Genome Biology | Full Text [https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02809-5]
• 72. A Phylogenomic Assessment of Processes Underpinning ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10161543/]
• 73. Phylogenomic Discordance is Driven by Wide-Spread ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11906154/]