The effects of purifying selection on patterns of genetic differentiation between Drosophila melanogaster populations.

Using the data provided by the Drosophila Population Genomics Project, we investigate factors that affect the genetic differentiation between Rwandan and French populations of D. melanogaster. By examining within-population polymorphisms, we show that sites in long introns (especially those >2000 bp) have significantly lower π (nucleotide diversity) and more low-frequency variants (as measured by Tajima's D, minor allele frequencies, and prevalence of variants that are private to one of the two populations) than short introns, suggesting a positive relationship between intron length and selective constraint. A similar analysis of protein-coding polymorphisms shows that 0-fold (degenerate) sites in more conserved genes are under stronger purifying selection than those in less conserved genes. There is limited evidence that selection on codon bias has an effect on differentiation (as measured by FST) at 4-fold (degenerate) sites, and 4-fold sites and sites in 8-30 bp of short introns ⩽65 bp have comparable FST values. Consistent with the expected effect of purifying selection, sites in long introns and 0-fold sites in conserved genes are less differentiated than those in short introns and less conserved genes, respectively. Genes in non-crossover regions (for example, the fourth chromosome) have very high FST values at both 0-fold and 4-fold degenerate sites, which is probably because of the large reduction in within-population diversity caused by tight linkage between many selected sites. Our analyses also reveal subtle statistical properties of FST, which arise when information from multiple single nucleotide polymorphisms is combined and can lead to the masking of important signals of selection.