Efficient Moment-Based Inference of Admixture Parameters and Sources of Gene Flow

Mark Lipson, Po‐Ru Loh, A. Levin, David Reich, Nick Patterson, Bonnie Berger

TL;DR

The authors introduce MixMapper, a semi-automated tool that builds phylogenetic trees incorporating "admixture" (gene flow between previously diverged populations). By using efficient moment-based statistics, it reveals that nearly all modern Europeans—including previously thought "isolated" groups like Basques and Sardinians—carry 20–40% ancestry from an ancient Northern Eurasian population.

What Problem Does This Paper Address?

Traditional phylogenetic trees are "branching only" models; they assume that once two populations diverge, they never meet again. However, human history is a messy story of migrations and interbreeding.

Existing tools to model this "admixture" fall into two camps:

  1. Likelihood-based methods: Highly precise but computationally "heavy," meaning they can only handle a few populations at a time and usually require the researcher to guess the tree structure (topology) beforehand.
  2. Fully automated moment-based methods (like TreeMix): Fast enough for large datasets but can sometimes produce "spurious" results or miss subtle signals because they try to solve the entire tree at once.

The authors sought a "middle ground": a tool that is computationally efficient, provides statistical confidence intervals, and allows for interactive, precise modeling of specific populations of interest.

Methods

MixMapper uses a novel two-phase approach based on $f$-statistics (measures of allele frequency correlations):

  • Phase 1: The Scaffold. The software identifies a subset of populations that show no signs of recent mixing. It uses these to build a "scaffold tree" via neighbor-joining. This provides a stable, unadmixed backbone.
  • Phase 2: Admixture Fitting. It then takes "target" populations and attempts to fit them onto the scaffold as a mixture of two (or three) different branches. It solves systems of linear equations to find the best-fit mixture proportions and the exact points on the branches where the ancestors diverged.
  • Moment Statistics: Instead of complex probability distributions, it uses the means and variances of allele frequency differences ($f_2$ and $f_3$ statistics). This makes the math much faster.
  • Drift Units: The paper introduces a new formula to convert abstract genetic distance into "drift units," which are more interpretable for biologists as they relate directly to population size and time.
  • Bootstrap Resampling: To ensure results aren't just noise, the tool runs hundreds of simulations (bootstraps) to provide confidence intervals for every parameter.

Datasets

  • Simulated Data: Generated using the ms coalescent simulator to test the algorithm against "ground truth" scenarios, including complex continuous gene flow.
  • HGDP (Human Genome Diversity Project): 934 individuals from 53 global populations.
  • Affymetrix Axiom Human Origins Array: A specialized SNP array designed to avoid "ascertainment bias" (the tendency for genetic tools to be biased toward the populations they were originally designed to study, usually Europeans).

Key Insights and Ideas

  • The "Universal" European Admixture: The most striking finding is that every European population sampled is best modeled as a mixture. They all share a signature of "Ancient Northern Eurasian" ancestry (related to the ancestors of Native Americans) and "Ancient Western Eurasian" ancestry.
  • Sardinians and Basques are not "Pure": These groups are often used in genetics as "unadmixed" references because they are outliers. MixMapper proves they are also admixed, just with more "drift" (genetic isolation) after the mixing event, which masked the signal for simpler tools.
  • Interactive Modeling: The authors argue that "semi-automation" is a feature, not a bug. By letting a researcher pick a high-quality scaffold, the tool avoids the errors that fully automated "black box" algorithms make when dealing with complex histories.

Key Results

  • European Ancestry: Inferred that Europeans have 20–40% Ancient Northern Eurasian ancestry.
  • Accuracy in Simulations: In complex scenarios where the popular tool TreeMix failed to identify the correct number or source of migration events, MixMapper recovered the true history with 100% confidence.
  • Computational Efficiency: After an initial setup of about an hour, MixMapper can fit a new admixed population in just a few seconds, allowing for real-time hypothesis testing.
  • Robustness: The results remained consistent even when the "scaffold" populations were changed, proving the model isn't overly sensitive to the initial setup.

Limitations and Future Work

  • Instantaneous Admixture: The model assumes admixture happens in a single "pulse." In reality, populations often migrate and mix continuously over centuries. The authors acknowledge this is a simplification.
  • Three-Way Limit: The current version is optimized for two-way and three-way mixtures. Modeling four or more ancestral sources would require significantly more data and computational complexity to avoid overfitting.
  • Scaffold Dependence: You need at least some unadmixed (or very anciently admixed) populations to build the scaffold. If every population in a dataset is recently and heavily mixed, the method struggles.
  • Archaic Admixture: The study intentionally ignored gene flow from archaic humans (like Neanderthals), which is a known but small component of the human genome.

Reference: Lipson, M., Loh, P. R., Levin, A., Reich, D., Patterson, N., & Berger, B. (2013). Efficient moment-based inference of admixture parameters and sources of gene flow. arXiv:1212.2555v2 [q-bio.PE].

Summary updated

Abstract

The recent explosion in available genetic data has led to significant advances in understanding the demographic histories of and relationships among human populations. It is still a challenge, however, to infer reliable parameter values for complicated models involving many populations. Here, we present MixMapper, an efficient, interactive method for constructing phylogenetic trees including admixture events using single nucleotide polymorphism (SNP) genotype data. MixMapper implements a novel two-phase approach to admixture inference using moment statistics, first building an unadmixed scaffold tree and then adding admixed populations by solving systems of equations that express allele frequency divergences in terms of mixture parameters. Importantly, all features of the model, including topology, sources of gene flow, branch lengths, and mixture proportions, are optimized automatically from the data and include estimates of statistical uncertainty. MixMapper also uses a new method to express branch lengths in easily interpretable drift units. We apply MixMapper to recently published data for Human Genome Diversity Cell Line Panel individuals genotyped on a SNP array designed especially for use in population genetics studies, obtaining confident results for 30 populations, 20 of them admixed. Notably, we confirm a signal of ancient admixture in European populations-including previously undetected admixture in Sardinians and Basques-involving a proportion of 20-40% ancient northern Eurasian ancestry.
BibTeX
@inproceedings{lipson2013efficient,
  author = {Mark Lipson and Po‐Ru Loh and A. Levin and David Reich and Nick Patterson and Bonnie Berger},
  title = {{Efficient Moment-Based Inference of Admixture Parameters and Sources of Gene Flow}},
  booktitle = {Molecular},
  year = {2013},
  doi = {10.1093/molbev/mst099},
}

Loading related content...

Figure