Adaptation and speciation: what can Fst tell us?

Mark BeaumontTrends 2005 · DOI

TL;DR

This paper revisits the Lewontin-Krakauer (LK) method, a statistical tool for identifying genes under natural selection by analyzing genetic variation (Fst) across populations. It argues that the LK method, when properly applied, is surprisingly robust to demographic history and offers a powerful first step in understanding adaptation and speciation, even in the face of gene flow.

What Problem Does This Paper Address?

For decades, evolutionary biologists have sought to understand how populations adapt to their environments and how new species arise. A key challenge is pinpointing the specific genes responsible for these evolutionary processes. While genetic variation within and between populations can provide clues, distinguishing the signal of natural selection from the noise of random genetic drift and complex demographic histories has been difficult. Early methods, like the original Lewontin-Krakauer (LK) test, were powerful but were criticized for relying on strong assumptions that often didn't hold in real-world scenarios, leading to their disuse. This paper aims to revive and re-evaluate the LK method, demonstrating its continued relevance and utility in modern genomics.

Methods

The paper focuses on the Lewontin-Krakauer (LK) method, which uses the inbreeding coefficient (Fst) to measure genetic differentiation between populations. The core idea is that loci experiencing natural selection (either favoring different alleles in different populations or undergoing balancing selection) will show different patterns of Fst compared to neutral loci.

The paper discusses:

  • The original LK test: This involved analyzing the variance of Fst estimates across multiple loci. A significant deviation from expected neutral variance suggested selection.
  • Criticisms of the original LK test: These included reliance on analytical approximations, assumptions about migration patterns, and the impact of mutation rates.
  • Modern re-evaluations: The paper highlights how computer simulations have overcome the limitations of analytical approximations, allowing for more realistic modeling of demographic scenarios and the distribution of Fst.
  • Bayesian approaches: Recent developments, such as those by Beaumont and Balding, use Bayesian frameworks to detect loci subject to selection, often incorporating a "separation-of-timescales" model. This model conceptually divides a population's genealogy into a recent "scattering phase" (within demes) and an ancestral "collecting phase" (metapopulation).
  • Alternative multilocus simulation-based tests: These methods, like those proposed by Schlotterer and colleagues, focus on "selective sweeps" and utilize different summary statistics beyond Fst.

The paper emphasizes that while there are now multiple variations of the LK approach, the underlying principle of using Fst distribution remains a valuable starting point.

Datasets

The paper discusses the application of Fst-based methods to various datasets, including:

  • Human populations: Early studies used data from worldwide human populations and the Yanomama tribe.
  • Snail (Cepaea nemoralis) and Peppered Moth (Biston betularia) populations: Used as examples of local adaptation.
  • Atlantic cod (Gadus morhua): Studies using RFLPs and allozymes to investigate population structure and selection.
  • Rodent genus Peromyscus: Including studies on the deermouse (Peromyscus maniculatus) to identify loci under selection related to altitude and anoxia.
  • Periwinkle (Littorina saxatilis): A study using AFLP markers to investigate differentiation in parapatric populations.

The paper does not introduce new benchmark datasets but rather reviews existing studies that have utilized these types of genetic data.

Key Insights and Ideas

  • Robustness to Demography: A central argument is that the distribution of Fst estimates is surprisingly robust to demographic history, especially when using the "separation-of-timescales" model. This means we can often detect selection without needing to precisely model complex population movements.
  • Fst as a First Step: The LK method, in its modern forms, is presented as a powerful and efficient first step for identifying candidate genes under selection, rather than a definitive endpoint.
  • Beyond Biochemical Function: The utility of Fst extends beyond identifying genes for biochemical function; it can help answer fundamental evolutionary questions about adaptation and speciation.
  • Speciation as a Genic Process: The paper suggests that speciation can be viewed from a "genic perspective," where specific genomic regions under strong disruptive selection or assortative mating effectively act as distinct species, even within a larger population.
  • Reduced Recombination and Adaptation: The importance of reduced recombination (e.g., due to inversions) in promoting adaptation in the face of gene flow is highlighted, as it can amplify the effects of selection across linked loci.

Key Results

  • The paper argues that the Lewontin-Krakauer method, when re-evaluated with modern computational tools, remains a valid and useful approach for identifying loci subject to selection.
  • Studies on Atlantic cod suggested that allozymes might be under balancing selection, while RFLPs were neutral, leading to a reassessment of population structure.
  • Research on Peromyscus identified the albumin locus as a strong candidate for adaptive differentiation, potentially linked to altitude and anoxia.
  • A study on Littorina saxatilis showed strong differentiation at specific markers, suggesting localized strong selection.
  • The paper implies that modern Fst-based surveys can effectively highlight genomic regions involved in disruptive selection and assortative mating, contributing to our understanding of speciation.

Limitations and Future Work

  • Ascertainment Bias: The paper acknowledges that markers with low mutation rates (like SNPs) can suffer from ascertainment bias if they were initially identified from a biased subset of populations.
  • High Mutation Rate Loci: Loci with high mutation rates (e.g., microsatellites) can be problematic as mutations may occur during the "scattering phase," violating the separation-of-timescales assumption.
  • Pairwise Comparisons: While restricting analysis to pairwise comparisons can overcome some issues, it might lead to a loss of statistical power and an increased risk of false positives.
  • Future Work: The authors suggest that further exploration of sampling strategies to maximize power and minimize false positives is needed. They also point to the long-term goal of statistically modeling selection at each locus alongside demographic history, rather than just identifying outliers. The paper also calls for more detailed theoretical work to understand the differences between various modern LK-based approaches.

Summary updated

BibTeX
@inproceedings{beaumont2005adaptation,
  author = {Mark Beaumont},
  title = {{Adaptation and speciation: what can Fst tell us?}},
  booktitle = {Trends},
  year = {2005},
  doi = {10.1016/j.tree.2005.05.017},
}
Figure