Using Patterns of Signed Linkage Disequilibria to Test for Epistasis in Flies and Plants

Whether deleterious mutations affect fitness independently, or synergistically, remains an open question in evolutionary genetics. Previous work by Sohail et al. (2017) reported an abundance of negative linkage disequilibrium (LD) values among loss-of - function (LOF) mutations in several human and fruit fly datasets, a pattern the authors interpreted as evidence of negative synergistic epistasis. Here we re-visit this question in a population genomic dataset of plants (Capsella grandiflora), and a fruit fly (Drosophila melanogaster) dataset previously used by Sohail et al. When using synonymous sites as a control, as Sohail et al., we find that both species have significantly less positive LD at LOF sites than synonymous sites. However, LD is not significantly different from 0 for LOF mutations but is significantly positive for synonymous mutations in both species. We question the use of synonymous sites as an appropriate control when attempting to make inferences about LD at selected sites. We use simulations to show how admixture or mating bias towards physically proximal individuals can cause positive LD to build up among neutral mutations but has a much weaker effect on selected sites, regardless of the presence of epistasis. Finally, we use information from published biological networks to explore whether there is evidence for negative synergistic epistasis between interacting radical missense mutations. We report no significant enrichment for negative inter-network LD in C. grandiflora. However, we note a modest but significant enrichment of negative LD in D. melanogaster networks, suggestive of intra-network negative synergistic epistasis in this species.