This guest post is by Nandita R. Garud, Philipp W. Messer, Erkan O. Buzbas, and Dmitri A. Petrov, on their paper
 Soft selective sweeps are the primary mode of recent adaptation in Drosophila melanogaster, arXived here

We typically think of adaptive events as arising from single de novo mutations that sweep through the population one at a time. In this scenario, one expects to observe the signatures of hard selective sweeps, where a single haplotype rises to very high frequencies, removing variation in linked genomic regions. It is also possible, however, that adaptation could lead to signatures of soft sweeps. Soft sweeps are generated by multiple adaptive haplotypes rising in frequency at the same time, either because (i) the adaptive mutation comes from standing variation and thus had time to recombine onto multiple haplotypes, or (ii) because multiple de novo mutations arise virtually simultaneously. The second mode is likely in large populations or when the adaptive mutation rate per locus is high.

Soft sweeps have generally been considered a mere curiosity and most scans for adaptation focus on the hard sweep scenario. Despite this prevailing view, the three best-studied cases of adaptation in Drosophila at the loci Ace, CHKov1, and Cyp6g1 all show signatures of soft sweeps. In two cases (Ace and Cyp6g1), soft sweeps were generated by de novo mutations indicating that the population size in D. melanogaster relevant to adaptation is on the order of billions or larger. In one case (CHKov1), soft sweeps arose from standing variation. Surprisingly, we do not have very convincing cases of recent adaptation in Drosophila that generated hard sweeps.

Nevertheless, it remained an open question of whether these three cases were the exception or the norm. They are all related to pesticide or viral resistance and it is entirely possible that much adaptation unrelated to human disturbance or immunity proceeds differently and might generate hard sweeps.

In this paper, we developed two haplotype statistics that allowed us to systematically identify hard and soft sweeps with similar power and then to differentiate them from each other. We applied these statistics to the Drosophila polymorphism data of ~150 fully sequenced, inbred strains available through the Drosophila Genetic Reference Panel (DGRP).

We found abundant signatures of recent and strong sweeps in the Drosophila genome with haplotype structure often extending over tens or even hundreds of kb. However, to our surprise, when we looked at the top 50 peaks, all of them showed signatures of soft sweeps, while we could not convincingly demonstrate the existence of any hard sweeps.

Our results suggest that hard sweeps might be exceedingly rare in Drosophila. Instead, it appears that adaptation in Drosophila primarily proceeds via soft sweeps and thus often involves standing genetic variation or recurrent de novo mutations. There are two caveats, however: One is that we were only able to study strong and recent adaptation. Such strong adaptation should “feel” recent population sizes that are close to the census size, whereas it should be insensitive to bottlenecks that have occurred in the distant past. Weaker adaptation, on the other hand, might take longer and thus would be sensitive to ancient bottlenecks or interference from other sweeps. Whether weak adaptation thus proceeds via hard sweeps remains to be seen. The second caveat is that much of adaptation might involve sweeps that are so soft and move so many haplotypes up in frequency that we cannot detect them. Similarly, adaptation could often be polygenic involving very subtle shifts in allele frequency at many loci. These modes would hardly leave any signatures of sweeps at all. Whichever way it is, it is becoming increasingly clear that adaptation in Drosophila and many other organisms is likely to be much more complex, much more common, and in many ways a much more turbulent process than we usually tend to think.