The HKA test revisited: a maximum-likelihood-ratio test of the standard neutral model.

We present a maximum-likelihood-ratio test of the standard neutral model, using multilocus data on polymorphism within species and divergence between species. The model is based on the Hudson-Kreitman-Aguade (HKA) test, but allows for an explicit test of selection at individual loci in a multilocus framework. We use coalescent simulations to show that the likelihood-ratio test statistic is conservative, particularly when the assumption of no recombination is violated. Application of the method to polymorphism data from 18 loci from a population of Arabidopsis lyrata provides significant evidence for a balanced polymorphism at a candidate locus thought to be linked to the centromere. The method is also applied to polymorphism data in maize, providing support for the hypothesis of directional selection on genes in the starch pathway.