Ancestral inference in population genetics models with selection
by
Peter Donnelly
University of Oxford
Coauthors: Matthew Stephens (University of Washington, Seattle)

A new algorithm is presented for exact simulation from the conditional distribution of the genealogical history of a sample, given the composition of the sample, for population genetics models with general diploid selection. The method applies to the usual diffusion approximation of evolution at a single locus, in a randomly mating population of constant size, for mutation models in which the distribution of the type of a mutant does not depend on the type of the progenitor allele; this includes any model with only two alleles. The new method is applied to ancestral inference for the two allele case, both with genic selection and heterozygote advantage and disadvantage, where one of the alleles is assumed to have resulted from a unique mutation event. We also describe how the method could be used for inference when data are available at neutral markers linked to the locus under selection.

Date received: June 2, 2002