In small populations, adaptation to the environment is likely limited by the availability of advantageous alleles. This is because in groups with small effective population size the low effective rate of new mutations and the low levels of segregating variation limit the probability of beneficial alleles to be present, at the right time and place, in the population. This affects positive selection both on new and on neutral standing variation, and both with monogenic and polygenic adaptation.
Variants under balancing selection hold the potential to mediate fast adaptations, if environmental change drives a shift from balancing to positive selection. This is because balanced alleles necessarily affect phenotype and fitness, and can be maintained by selection even in small populations. Other non-neutral variants can also be maintained if they are tightly linked to sites under balancing selection. We discuss this model of positive selection on previously balanced alleles and show, using forward simulations, that it has high potential to mediate fast, local adaptation in human populations. We also introduce a statistic (DIFFSS) that jointly considers the site frequency spectra of two populations and has high power to identify positive selection on loci previously under balancing selection. We applied DIFFSS on the genomes of African and Eurasian human populations and, after combining these results with Approximate Bayesian Computations, identified dozens of loci with signatures of this mode of adaptation.