| A key question in population genetics is the extent to which positive selection drives molecular evolution. According to the selectionist viewpoint, evolution at the molecular level occurs by natural selection acting on DNA sequence mutations, with selectively favorable mutations more likely to eventually reach fixation in a species. On the other hand, the neutral theory of evolution postulates that random genetic drift, not selection, is the major driving force behind evolution at the molecular level.;Here, we address this question within a Poisson random field framework, based on aligned DNA sequence data from two closely related species. We investigate heavy-tailed distributions for within-locus selection coefficients, specifically a double-exponential and a Student's t distribution. Using Markov chain Monte Carlo methods on a set of coding sequences from 91 autosomal genes in Drosophila melanogaster and Drosophila simulons, we estimate that while few (∼ 10%) new mutations are beneficial, many 40%) of observed polymorphic sites and most (∼ 95%) fixations are due to positive selection. While these results are necessarily model-dependent, they are in accord with previous estimates based on a normal random effects model.;As a second aim, we develop programs for forward simulation of polymorphism and divergence data under varying levels of selection and recombination to investigate the performance of the Poisson random field model when the assumption of independence between polymorphic sites is not met. We find slightly diminished performance of the model under tight linkage and examine the reasons why this may be so. |
