A New Method to Uncover Signatures of Divergent and Stabilizing Selection in Quantitative Traits

Abstract

While it is well understood that the pace of evolution depends on the interplay between natural selection, random genetic drift, mutation, and gene flow, it is not always easy to disentangle the relative roles of these factors with data from natural populations. One popular approach to infer whether the observed degree of population differentiation has been influenced by local adaptation is the comparison of neutral marker gene differentiation (as reflected in F ST ) and quantitative trait divergence (as reflected in Q ST ). However, this method may lead to compromised statistical power, because F ST and Q ST are summary statistics which neglect information on specific pairs of populations, and because current multivariate tests of neutrality involve an averaging procedure over the traits. Further, most F ST -Q ST comparisons actually replace Q ST by its expectation over the evolutionary process and are thus theoretically flawed. To overcome these caveats, we derived the statistical distribution of population means generated by random genetic drift and used the probability density of this distribution to test whether the observed pattern could be generated by drift alone. We show that our method can differentiate between genetic drift and selection as a cause of population differentiation even in cases with F ST ¼ Q ST and demonstrate with simulated data that it disentangles drift from selection more accurately than conventional F ST -Q ST tests especially when data sets are small. N ULL models have an important role in many areas of biology (e.g., Whitlock and Phillips 2000;Hubbell 2001;Ohta 2002) and not least in studies of population differentiation. For instance, the expectation that genetic differentiation occurs most readily toward the direction of maximum additive genetic variance (Schluter 1996) has provided a useful null model for evaluating the role of the ancestral G matrix in the rate and direction of evolutionary transitions (e.g., McGuigan 2006). Likewise, while few evolutionary biologists would doubt the power of natural selection over genetic drift (Hohenlohe and Arnold 2008), comparisons of the index of quantitative genetic differentiation (Q ST ) to the index of neutral genetic differentiation (F ST ) have provided a useful platform to identify traits and populations subject to directional selection (Merilä and Crnokrak 2001;McKay and Latta 2002;Leinonen et al. 2008).The current neutrality tests of population differentiation make restrictive assumptions or suffer from various methodological or logical problems (e.g., O'Hara and Merilä 2005;Beaumont 2008; Whitlock 2008). To start with, the popular method of comparing the index of quantitative genetic differentiation (Q ST ) to the index of neutral genetic differentiation (F ST ) does not necessarily provide practical means to analyze whether a small number of populations are locally adapted to their environments, as obtaining reliable estimates typically requires data from .10 populations in controlled environmental conditions (O'Hara ...

Contrasting the distribution of phenotypic and molecular variation in the freshwater snail Biomphalaria pfeifferi, the intermediate host of Schistosoma mansoni

“…It is unlikely that this holds (F ST estimates are whatsoever pretty high) because, even in highly selfing species, highly variable microsatellites would produce much higher within-population variability than observed here (compare with for example, Viard et al, 1997). Both Whitlock and Guillaume (2009) and Ovaskainen et al (2011) raised statistical issues when comparing Q ST and F ST . The method of Whitlock and Guillaume (2009) is difficult to implement as typical data sets (as ours) rarely include enough loci to directly infer the distribution of F ST without extra inferential steps, and Q ST for a trait is rarely measured with high precision, so the position of a given estimated Q ST value in the distribution cannot be known without error.…”

“…The method of Whitlock and Guillaume (2009) is difficult to implement as typical data sets (as ours) rarely include enough loci to directly infer the distribution of F ST without extra inferential steps, and Q ST for a trait is rarely measured with high precision, so the position of a given estimated Q ST value in the distribution cannot be known without error. The approach promoted by Ovaskainen et al (2011) seems to be the most appropriate at this point, but is unfortunately too intricate to be conducted in a convenient way.…”

“…A basic prediction is that diversifying selection can be inferred when F ST oQ ST, whereas stabilizing selection is at play when F ST 4Q ST . This approach has been criticized, especially with regard to the latter inference, because the respective values of F ST and Q ST depend on genetic architecture (Goudet and Büchi, 2006;Pujol et al, 2008), relative genetic variability at neutral markers and traits , and for statistical reasons (Whitlock and Guillaume, 2009;Ovaskainen et al, 2011). It remains though useful as a first approach.…”

Selection on the morphology–physiology‐performance nexus: Lessons from freshwater stickleback morphs

“…To ascertain whether natural selection or random genetic drift was responsible for trait divergence, we employed a model of multivariate trait differentiation expected under neutrality (Ovaskainen, Karhunen, Zheng, Arias, & Merilä, 2011), as implemented in the package “driftsel” (Karhunen, Merilä, Leinonen, Cano, & Ovaskainen, 2013). In this model, differentiation from a common ancestral population expected under drift is derived from both neutral genetic markers and the pedigrees of experimental animals.…”

“…The signature of selection is reflected in the model parameter S : under neutral divergence the expected values of S is 0.5, whereas values <0.5 indicate stabilizing selection and those >0.5 reflect divergent selection (Ovaskainen et al., 2011). To determine the significance of S , we profiled 95% PDIs and p ‐values from the posterior distribution of the 1,000 estimates derived from the sampling chain.…”