Detecting Directional Selection from the Polymorphism Frequency Spectrum

摘要: The distribution of genetic polymorphisms in a population contains information about the mutation rate and strength natural selection at locus. Here, we show that Poisson Random Field (PRF) method population-genetic inference suffers from systematic biases tend to underestimate pressures rates, erroneously infer positive selection. These problems arise infinite-sites approximation inherent PRF method. We introduce three new techniques correct these problems. present finite-site modification method, as well two methods for inferring rates based on diffusion models. Our can be used not only "weighted average" acting gene sequence, but also across sites. evaluate accuracy our methods, original approach, by comparison with Wright-Fisher simulations.