Adaptation and speciation: what can F_st tell us?

A useful way of summarizing genetic variability among different populations is through estimates of the inbreeding coefficient, F_st. Several recent studies have tried to use the distribution of estimates of F_st from individual genetic loci to detect the effects of natural selection. However, the promise of this approach has yet to be fully realized owing to the pervasive dogma that this distribution is highly dependent on demographic history. Here, I review recent theoretical results that indicate that the distribution of estimates of F_st is generally expected to be robust to the vagaries of demographic history. I suggest that analyses based on it provide a useful first step for identifying candidate genes that might be under selection, and explore the ways in which this information can be used in ecological and evolutionary studies.

Introduction

It is becoming increasingly cheaper and faster to survey samples genetically from model and non-model organisms for a large number of loci across their genomes, an advance that derives largely from the activities of the biomedical community [1]. Here, I reappraise an old idea [2] for the analysis of such data, first proposed when multi-locus surveys were scarce and difficult to obtain; it was later discredited 3, 4 and abandoned; and now, fuelled by the plentiful supply of these data, occasionally peeps apologetically out of research articles, smothered in caveats [5].

The ready availability of different classes of gene frequency information has rekindled an interest in natural selection and the development of a variety of methods for use in trying to infer the presence and mode of selection. Three main approaches can be identified [6]: (i) detailed modelling of selection at individual loci or sequences; (ii) multilocus comparisons, of which the Lewontin–Krakauer Method (see Glossary [2]), discussed here, is the oldest; and (iii) comparison of patterns of substitution among synonymous and non-synonymous sites. Analyses of the advantages and drawbacks of these different approaches are detailed by Nielsen [6]. Much of the research has been driven by the biomedical community with an aim to identify and characterize biochemical function and the phenotypic effect of natural variation throughout the human genome, often based on comparative analysis [7]. Judgements about the efficacies of different methods implicitly tend to have these, ultimately medical, goals in mind.

In evolutionary biology there are, by contrast, several interesting hypotheses that can be tested by characterizing the number, position and fitness effects of genomic regions that show apparent adaptive divergence in allele frequency, without the need to delve into the physiological details. I argue that, for most organisms, the easiest way to achieve this is by using the Lewontin–Krakauer method, which appears, at least in recent versions, to be generally robust to the vagaries of demographic history. I invoke recent theoretical results that suggest why this is not so surprising and discuss possible sampling strategies that might maximize the power of the approach. I also outline areas of application, particularly the study of adaptive divergence in the face of gene flow and modes of speciation.