local adaptation

Fariello MI, Boitard S, Naya H, SanCristobal M & Servin B 2013 Detecting signatures of selection through haplotype differentiation among hierarchically structured populations. Genetics 193:929-941.

  • Bonhomme et al. (2010) proposed an extension of the classical Lewontin and Krakauer (LK) test (Lewontin and Krakauer 1973), where the hierarchical population structure is captured through a kinship matrix, which is used to model the covariance matrix of the population allele frequencies
  • a similar covariance matrix was also introduced in a related context to account for the correlation structure arising from population geography
  • for the detection of selection within species (the ecological scale of time), methods can be classified into three groups:
  • methods based on (i) the high frequency of derived alleles and other consequences of hitchhiking within population
  • (ii) the length and structure of haplotypes, measured by extended haplotype homozygosity (EHH) or EHH-derived statistics
  • (iii) the genetic differentiation between populations, measured by FST or related statistics
  • methods of the latter kind, which we focus on, are particularly suited to the study of species that are structured in well-defined populations
  • in contrast to methods based on the frequency spectrum (i) or the excess of long haplotypes (ii), they can detect a wider range of selection scenarios, including selection on standing variation or incomplete sweeps
  • in the case of sequencing data, we found that a single SNP test was more powerful than hapFLK, consistent with previous results of Innan and Kim (2008), who found in a similar setting that an haplotype-based FST was less powerful than a single locus one
  • in many real situations, selection will act rather effectively on multilocus haplotypes (Pritchard et al. 2010), due, for instance, to recurrent mutations affecting the same gene, or to polygenic selection
  • we expect haplotype-based tests to be more powerful in such situations, which according to us justifies their use also for the analysis of sequencing data
  • soft or incomplete sweeps
  • genome scans for selection have historically focused on hard sweeps
  • several recent studies have pointed out the importance of soft sweeps in the evolution of populations
  • we tested hapFLK for initial frequencies of the favorable allele up to 30% and found that reasonable power could be achieved also in this situation
  • the detection of incomplete sweeps is another important issue, which has not been much tackled in the literature
  • detecting selected alleles at intermediate frequency is almost impossible with methods based on the allele frequency spectrum and very difficult with EHH- or FST-based existing approaches
  • hapFLK is quite powerful in the case of incomplete sweeps
  • several of the selection signatures detected in the sheep HapMap data correspond to intermediate frequencies of the selected haplotype
  • few hard sweeps were actually detected in the sheep data
  • this might be due to the short divergence time between these populations (a few hundred generations)
  • our study demonstrates that using haplotype information in FST-based tests for selection greatly increases their detection power
  • it also confirms the importance of analyzing multiple populations jointly