Abstract
Natural selection, in the form of balancing selection or selective sweeps, can result in a decoupling of the amounts of molecular polymorphism and divergence. Thus natural selection can cause some areas of DNA sequence to have greater silent polymorphism, relative to divergence between species, than other areas. It would be useful to have a statistical test for heterogeneity in the polymorphism to divergence ratio across a region of DNA sequence, one that could identify heterogeneity greater than that expected from the neutral processes of mutation, drift, and recombination. The only currently available test requires that a region be arbitrarily divided into sections that are compared with each other, and the subjectivity of this division could be problematic. Here a test is proposed in which runs of polymorphic and fixed sites are counted, where a "run" is a set of one or more sites of one type preceded and followed by the other type. The number of runs is smaller than otherwise expected if polymorphisms are clumped together. By simulating neutral evolution and comparing the observed number of runs to the simulations, a statistical test is possible which does not require any a priori decisions about subdivision.
