Inferring selection strength under complex demographic scenarios

The strength of natural selection against transposable elements (TEs) can be inferred from the frequencies of a sample of TEs. However, complicated demographic histories could lead to a substantial distortion of the TE frequency distribution compared to that expected for a panmictic, constant-sized population. In a paper just published by MBE Josefa Gonzalez, Mike Macpherson, Philip Messer, and Dmitri develop a flexible maximum likelihood methodology that explicitly accounts both for demographic history and for the ascertainment biases of identifying TEs. We apply this method to the newly generated frequency data of the BS family of non-LTR retrotransposons in D. melanogaster in concert with two recent models of the demographic history of the species to infer the intensity of selection against this family. We find the estimate to differ substantially from our own prior estimates made under the assumptions of panmixis. These findings highlight the importance of accounting for demographic history and bear on study design for the inference of selection coefficients generally.