Rogers et al. present a detailed analysis of tandem duplications in the Drosophila simulans and D. yakuba genomes. They find that duplicates generally show a skew in the site frequency spectrum (SFS) compared to putatively neutral SNPs (not neutral sites – thanks Casey!) in introns. On the autosomes this shows up as an excess of rare variants, suggesting duplicates are generally deleterious. On the D. simulans X they find evidence of positive selection on duplicates, and analysis of GO terms seems consistent with this. Overall, the authors estimate that gene duplication mutations are rare, and their frequent removal from the population by purifying selection means there is very little standing genetic variation for tandem duplicates. This means that populations will be mutation limited if a duplication is required for adaptation.

One place I would have liked to see more discussion is on the genetic architecture of traits. One might imagine that for highly quantitative traits populations don’t need to wait for gene duplications, as the mutational target may be quite large and there is likely to be considerable standing genetic variation. For such traits, might we not expect to see repeated evolution (the authors use both “parallel” and “convergent” interchangeably, but I find both terms somewhat confusing and prefer the simpler “repeated”)? But for traits with simpler architectures and a smaller mutational target might gene duplications be expected play a more important role? This idea seems to fit with some of their data – traits where repeated evolution is seen in both taxa include traits such as immunity that may have simple genetic architectures. The authors instead argue that gene duplications may be more important for more quickly evolving phenotypes, but it is not at all clear to me that highly quantitative traits can’t evolve quickly.

In their discussion of repeated evolution between D. simulans and D. yakuba the authors ignore duplicates fixed within each lineage. If the question is one of repeated evolution between these two taxa, why not include such changes?

In their assessment of the SFS of gene duplicates, the authors make comparisons to putatively neutral intronic sites SNPs and, upon finding differences, argue that gene duplicates are generally deleterious and of large effect. I would have liked to see comparison to nonsynonymous sites SNPs as well – do duplicates show a larger skew?

A few minor points merit mention. The discussion of “convergent” evolution is a bit confusing, as the authors argue in the introduction that it’s the best evidence for natural selection but in the discussion that it’s probably a bad metric. And when discussing the likelihood of adaptation from tandem duplicates the authors cite Haldane’s sieve a reason why recessive gene duplicates might be less likely to fix, but ignore the fact that adaptation from standing variation can change such dynamics.