Recombination and selection against introgressed DNA

Abstract

DNA introgressed from one species into another is typically deleterious at many genomic loci in the recipient species. It is therefore purged by selection over time. Here, we use mathematical modeling and whole-genome simulations to study the influence of recombination on the purging of introgressed DNA. We find that aggregate recombination controls the genome-wide rate of purging in the first few generations after admixture, when purging is most rapid. Aggregate recombination is quantified by [Formula], the average recombination rate across all locus pairs, and analogous metrics. It is influenced by the number of crossovers (i.e., the map length) and their locations along chromosomes, and by the number of chromosomes and heterogeneity in their size. A comparative prediction of our analysis is that species with fewer chromosomes should purge introgressed DNA more profoundly, and therefore should exhibit a weaker genomic signal of historical introgression. With regard to patterns across the genome, we show that, in heterogametic species with autosomal recombination in both sexes, more purging is expected on sex chromosomes than on autosomes, all else equal. The opposite prediction holds for species without autosomal recombination in the heterogametic sex. Finally, we show that positive genomic correlations between local recombination rate and introgressed ancestry, as recently observed in several taxa, are likely driven not by recombinations effect in unlinking neutral from deleterious introgressed alleles, but rather by its effect on the rate of purging of the deleterious alleles themselves. Note on this versionAn earlier version of this manuscript had two parts: (1) Calculations of the variance of genetic relatedness between individuals with particular pedigree relationships, taking into account the randomness of recombination and segregation in their pedigree. (2) An investigation of the rate of purging of introgressed DNA following admixture, based in part on results from part (1). Part (1) has since been published as Veller et al. (2020). The present manuscript has been reconfigured to focus on part (2).