A gene-based model of fitness and its implications for genetic variation

A widely used model of the effects of mutations on fitness (the ″sites″ model) assumes that heterozygous recessive or partially recessive deleterious mutations at different sites in a gene complement each other, similarly to mutations in different genes. However, the general lack of complementation between major effect allelic mutations suggests an alternative possibility, the ″gene″ model. This assumes that a pair of heterozygous deleterious mutations in trans behave effectively as homozygotes, so that the fitnesses of trans heterozygotes are lower than those of cis heterozygotes. We examine the properties of the two different models, using both analytical and simulation methods. We show that the gene model results in a slightly lower mutational load, but a much smaller inbreeding load, than the sites model, implying that standard predictions of mutational contributions to inbreeding depression may be overestimates. The gene model also predicts positive linkage disequilibrium (LD) between derived variants within the coding sequence under conditions when the sites model predicts zero or slightly negative LD. We also show that focussing on rare variants when examining patterns of LD, especially with Lewontin′s D′ measure, is likely to produce misleading results with respect to inferences concerning the causes of the sign of LD. Synergistic epistasis between pairs of mutations was also modeled; it is less likely to produce negative LD under the gene model than the sites model. The theoretical results are discussed in relation to data on inbreeding load in Drosophila melanogaster and patterns of LD in natural populations of several species.