A two-step method for detecting selection signatures using genetic markers

A two-step procedure is presented for analysis of theta (FST) statistics obtained for a battery of loci, which eventually leads to a clustered structure of values. The first step uses a simple Bayesian model for drawing samples from posterior distributions of theta-parameters, but without constructing Markov chains. This step assigns a weakly informative prior to allelic frequencies and does not make any assumptions about evolutionary models. The second step regards samples from these posterior distributions as 'data' and fits a sequence of finite mixture models, with the aim of identifying clusters of theta-statistics. Hopefully, these would reflect different types of processes and would assist in interpreting results. Procedures are illustrated with hypothetical data, and with published allelic frequency data for type II diabetes in three human populations, and for 12 isozyme loci in 12 populations of the argan tree in Morocco.