Estimating self-fertilization rates from the diploid genotype spectrum

Abstract

The distribution of self-fertilization across the tree of life remains poorly characterized. Here we present the diploid genotype spectrum (DGS) as a useful genomic data summary for characterizing populations undergoing partial self-fertilization (selfing). This statistic provides a straightforward baseline for estimating selfing rates from population genomic data. At every segregating site across a genome, the diploid genotype configuration records the numbers of homozygous ancestral, heterozygous, and homozygous derived individuals, and the DGS compiles the counts of these configurations across sites. We show that the expected relative frequencies of DGS configurations among segregating sites can be computed under the standard coalescent approximation for partial selfing. We present two methods for doing this, one that is an intuitive recursion through possible states of the model, and one that is a more efficient combinatorial version. We give a worked example, validate the analytical expectations using forward simulations, and introduce a software package that estimates population selfing rates from diploid SNP data. The computations here mechanistically link selfing to observed patterns of diploid genotype configurations.