In our manuscript exploring the population genetics of local adaptation (Tiffin and Ross-Ibarra 2014) we included a discussion about the potential uses of reduced representation data (e.g. RAD-seq, GBS). To provide a sense of the probability of using reduced representation data to identify targets of selection, we included a figure showing the probability of having a SNP included in a region of the genome in which diversity had been severely reduced due to a recent selective sweep. Unfortunately this figure is not correct; an error in the code inadvertently used centimorgans as morgans, causing the recombination rate to be off by a factor of 100.
To correct this we have generated a new figure (see http://rpubs.com/rossibarra/257207; raw code is available at https://gist.github.com/rossibarra/be44cc3b3796f45840d942ad11c01ba1) that corrects this error and presents a more realistic model. Our previous model assumed SNPs were distributed evenly across the genome and the presence of a single SNP near a sweep was sufficient for detection. Instead, here we explicitly model sequence “tags” coming from RAD-seq or GBS, and incorporate information about the variation in diversity expected among tags in neutral regions of the genome. The figure clearly shows that with dense marker coverage and strong selection, the probability of detecting reductions in diversity due to recent selective sweeps from new beneficial mutations can be relatively high. We emphasize, however, that the purpose of the figure is solely to develop an intuition of the likelihood of detecting a recent selective sweep. The many simplifying assumptions made in generating the figure (no recent demographic change, both sequence tags and recombination occur uniformly along the genome, selection is on a novel beneficial mutation with additive effect that has recently swept to fixation), as well as the specific mutation rates, sample size, sequence length, and recombination rates assumed will all affect the actual probability of a tag being included in a selective sweep. Moreover, this figure does not touch on many other relevant issue such as multiple testing, complex demography, background selection, or other modes of positive selection (e.g. from standing variation, balancing selection, or selection on polygenic traits).
We have submitted a correction to the journal.
We thank Eric Johnson for drawing our attention to the error, and Eric Johnson, Kathleen Lotterhos, and Graham Coop for kindly reviewing previous versions of the code and assumptions we have used in generating this new figure.