Detecting frequency-dependent selection through the effects of genotype similarity on fitness components

Abstract

Frequency-dependent selection (FDS) drives an evolutionary regime that maintains or disrupts polymorphisms. Despite the increasing feasibility of genetic association studies on fitness components, there are a few methods to uncover the loci underlying FDS. Based on a simplified model of pairwise genotype-genotype interactions, we propose a linear regression that can infer FDS from observed fitness. The key idea behind our method is the inclusion of genotype similarity as a pseudo-trait in selection gradient analysis. Single-locus analysis of Arabidopsis and damselfly data could detect known negative FDS on visible polymorphism that followed Mendelian inheritance with complete dominance. By extending the singlelocus analysis to genome-wide association study (GWAS), our simulations showed that the regression coefficient of genotype similarity can distinguish negative or positive FDS without confounding other forms of balancing selection. Field GWAS of the branch number further revealed that negative FDS, rather than positive FDS, was enriched among the top-scoring single nucleotide polymorphisms (SNPs) in Arabidopsis thaliana. These results showed the wide applicability of our method for FDS on both visible polymorphism and genome-wide SNPs. Our study provides an effective method for selection gradient analysis to understand the maintenance or loss of polymorphism.