Joint estimation of selection intensity and mutation rate under balancing selection with applications to HLA.

A composite likelihood method is introduced for jointly estimating the intensity of selection and the rate of mutation, both scaled by the effective population size, when there is balancing selection at a single multi-allelic locus in an isolated population at demographic equilibrium. The performance of the method is tested using simulated data. Average estimated mutation rates and selection intensities are close to the true values but there is considerable variation about the averages. Allowing for both population growth and population subdivision does not result in qualitative differences but the estimated mutation rates and selection intensities do not in general reflect the current effective population size. The method is applied to 3 class I (HLA-A, HLA-B, and HLA-C) and 2 class II loci (HLA-DRB1 and HLA-DQA1) in the 1000 Genomes populations. Allowing for asymmetric balancing selection has only a slight effect on the results from the symmetric model. Mutations that restore symmetry of the selection model are preferentially retained because of the tendency of natural selection to maximize average fitness. However, slight differences in selective effects result in much longer persistence time of some alleles. Trans-species polymorphism, which is characteristic of major-histocompatibility loci in vertebrates, is more likely when there are small differences in allelic fitness than when complete symmetry is assumed. Therefore, variation in allelic fitness expands the range of parameter values consistent with observations of trans-species polymorphism.