Introgression Detection - Usage Guide

Overview

Tests for inter-population admixture combine site-pattern statistics (ABBA-BABA, f4), tree-topology methods (Dsuite f-branch, QuIBL, Twisst), explicit networks (PhyloNet, qpGraph), and haplotype-based methods (sprime, Relate). The fundamental confounder is incomplete lineage sorting (ILS): a significant Patterson D-statistic is consistent with introgression OR with asymmetric ILS OR with ancestral structure OR with ghost lineages (Eriksson & Manica 2012 PNAS 109:13956).

For publication-grade introgression claims, combine D-statistic with Fbranch mapping (Malinsky 2018), Twisst / QuIBL topology weighting at the locus level, and at least one network method.

Prerequisites

# Modern Dsuite
git clone https://github.com/millanek/Dsuite && cd Dsuite && make
# Or: conda install -c bioconda dsuite

# Population-genetics ABBA-BABA + qpAdm
git clone https://github.com/DReichLab/AdmixTools && cd AdmixTools && make
# Modern AdmixTools v2 (R wrapper)
Rscript -e "remotes::install_github('uqrmaie1/admixtools')"

# TreeMix
conda install -c bioconda treemix

# Locus-level topology
git clone https://github.com/miriamtnzr/QuIBL  # QuIBL
git clone https://github.com/simonhmartin/twisst # Twisst

# Network methods
git clone https://github.com/NakhlehLab/PhyloNet && cd PhyloNet && mvn package

# Haplotype-based
conda install -c bioconda sprime

# Hybridization-specific
git clone https://github.com/pblischak/HyDe && cd HyDe && pip install -e .

# Optimal migration count
Rscript -e "install.packages('OptM')"

Quick Start

Tell the AI agent what to do: - "Compute Patterson's D-statistic between these populations using Dsuite Dtrios" - "Apply Fbranch to assign admixture signal to specific tree branches" - "Run TreeMix with migration edges and select optimal number with OptM" - "Use sprime to identify archaic introgressed haplotype tracts in modern human populations"

Example Prompts

Inter-Species Admixture Test

"Test for introgression between Heliconius melpomene and H. cydno using Dsuite Dtrios with H. erato as outgroup. Apply Fbranch mapping. Cross-validate with Twisst topology weighting in 50 kb windows across the genome to identify specific introgressed regions."

Population Admixture Modeling

"Test if the modern Argentinean population can be modeled as admixed from Spanish + Native American sources using qpAdm rotation. Cross-validate with TreeMix migration edges. Use OptM (Evanno method) to select optimal number of migration edges."

Archaic Introgression

"Identify Neanderthal introgressed haplotype tracts in modern non-African human samples using sprime. Verify that tract lengths cluster around expected size (>50 kb typical for archaic). Compare introgression rates across populations."

What the Agent Will Do

Validate inputs: VCF format; sample-population mapping; outgroup choice; population sample sizes
Compute D-statistic with Dsuite Dtrios across all population trios
Apply Fbranch to assign admixture to specific tree branches
Cross-validate locus level with Twisst (topology weighting per window) or QuIBL (per gene tree)
Build migration model with TreeMix; select migration count via OptM
qpAdm rotation tests for proposed source populations
For ancient admixture: qpGraph manual topology design
For archaic introgression: sprime per-individual analysis
Report: D + Z-score per trio, Fbranch attributions, Twisst topology weights, qpAdm rotation pass/fail
Caveats: ILS vs introgression, ghost lineages, ancestral structure, outgroup distance

Tips

Always interpret D > 0 as "consistent with introgression OR ILS OR ancestral structure" (Eriksson & Manica 2012)
Use Fbranch (Malinsky 2018) to assign signal to specific branches; sharper than raw D
Cross-validate with Twisst / QuIBL at locus level to distinguish ILS from introgression
For multi-population analysis, qpAdm + TreeMix combination is gold standard
TreeMix migration count selection: use OptM Evanno method, not raw likelihood
For divergent species, use ABBAclustering (Koppetsch-Malinsky-Matschiner 2024 Syst Biol)
Outgroup distance: < 100 Myr ideal; > 200 Myr introduces convergent-evolution artifacts
One individual per population is standard (Dsuite default); for population-genetic, 5+ samples
Always report sample sizes; small N inflates D variance
Multiple-testing correction across many trios: FDR or Bonferroni
For sprime, restrict to populations where archaic introgression source is known
For ancient admixture, qpGraph with explicit topology > automatic methods
ILS confound dominates short internodes; report internode length context
Cytonuclear discordance: mtDNA introgression may differ from autosomal

comparative-genomics/hgt-detection - Distinguish HGT from hybridization in microbes comparative-genomics/gene-tree-species-tree-reconciliation - Reconciliation alternative for introgression comparative-genomics/synteny-analysis - Per-window topology analysis population-genetics/population-structure - PCA / ADMIXTURE precedes introgression testing population-genetics/selection-statistics - Selection on introgressed regions phylogenetics/species-trees - Coalescent species tree variant-calling/joint-calling - Multi-sample VCF input read-alignment/bwa-alignment - Alignment underlies variant calling