polygenic adaptation

Ferrer-Admetlla A, Liang M, Korneliussen T & Nielsen R 2014 On detecting incomplete soft or hard selective sweeps using haplotype structure. Mol Biol Evol 31:1275-1291.

• multiple independent mutations at a single locus are all favored and increase in frequency simultaneously until the sum of the frequencies is 1 (polygenic adaptation)
• [?]

• given the currently accepted assumptions regarding human demography (a small effective population size and migration out of Africa 50–100 ka), there may have been little time for new beneficial mutations to occur
• the hard sweep model may not be entirely appropriate for describing the process of adaptation in recent human history

• selection on standing variation occurring on more than one loci (polygenic adaptation from standing variation) has been suggested to be an important, if not the most important, mechanism of adaptation in humans

• haplotype patterns will clearly change even when selection is acting on multiple haplotypes
• haplotype-based statistics are, therefore, an obvious avenue to pursue when designing methods for detecting sweeps from standing variation

• most scans for selection in the human genome had a strong bias toward identifying regions of low recombination
• the possibility that many results in fact are false positives due to reduced recombination is worrying

• the main difference between the nS_L and iHS statistics is in how they measure distance
• the nS_L statistic uses segregating sites as a proxy for distance
• the iHS statistic uses the recombination distance
• iHS can be viewed as nS_L with some additional randomness due to the spacing between segregating sites

• a program, and associated open source code for computing nS_L is available at http://cteg.berkeley.edu/~nielsen/

• the haplotype-based methods are known to have most power when the allele frequency is moderately high (65–85%)

• the window size determines the maximum allowed length of a haplotype
• in EHH and iHS, the use of a window size was originally introduced for purely computational reasons to reduce computational complexity
• even under a standard neutral model, the expected length of homozygosity segment is infinite
• the expected time to the next mutation or recombination in a fragment with initial coalescence time t between a pair of sequences, is given by [t(θ + 2ρ)]^{− 1}
• the integral of this expectation over the distribution of t in the standard neutral model does not converge