It is critical to have accurate estimates of mutational rates and biases in order to successfully perform many population genetic analyses. Yet, because a substantial fraction of new mutations are deleterious, our ability to detect de novo events is severely limited. Fortunately, the scientific community studying Drosophila melanogaster has generated a large body of sequence data that is publicly available. We leverage these resources, in combination with additional strains sequenced in our lab, to investigate mutational patterns in Drosophila melanogaster using two different strategies.
The first strategy is via mutational accumulation, an experiment that allows new mutations to accrue in the genomes of laboratory strains, and which has now been conducted three independent times. The two previously published experiments used the inbreeding method of mutation accumulation (Keightley et. al. 2009, Schrider et. al. 2013) and to this we are contributing a novel data set that was generated using an alternative experimental method, one which additionally permits recessive strongly deleterious mutations to accumulate. We find a significantly higher mutation rate than previously reported, suggesting that strongly deleterious recessive mutations are an important part of the mutational spectrum. In other metrics the three experiments have very similar patterns, and thus we merge these three resources and give estimates for a variety of fundamental mutational biases.
Mutation accumulation studies are excellent for measuring mutation rates, yet our ability to characterize other more subtle mutational patterns is still limited by the small number of events which occur during these experiments. To address this, we combine the individually sequenced strains made available by the Drosophila Genome Nexus (which includes lines from DGRP, DPGP, and DSPR) (Lack et. al. 2015), with newly generated sequence data using pooled individuals from natural populations, which collectively represent >17,000X coverage of >5,000 strains. These data sets allow the detection of extremely rare polymorphisms, which we further validated with resequencing. Extremely rare variants are a class of sites enriched for new deleterious mutations, because at low frequencies their dynamics are dominated by stochastic rather than selective forces. We show that the these extremely rare variants in fact approach the de novo spectrum as revealed by mutation accumulation experiments, and use this high quality set of rare variants to measure additional mutational patterns.