The dynamics of alternative pathways to compensatory substitution

TL;DR

This paper analytically derives and verifies the rates and probabilities of two pathways for compensatory substitutions under weak to moderate selection, exploring their dynamics and implications for phylogenetic inference.

Contribution

It introduces a comprehensive analytical framework for understanding two alternative pathways of compensatory substitution in population genetics under realistic selection regimes.

Findings

Rates of substitution depend on population size, mutation, and recombination.

Paths have similar durations under moderate selection, diverge under weak selection.

Recombination influences the dynamics of compensatory evolution.

Abstract

The role of epistatic interactions among loci is a central question in evolutionary biology and is increasingly relevant in the genomic age. While the population genetics of compensatory substitution have received considerable attention, most studies have focused on the case when natural selection is very strong against deleterious intermediates. In the biologically-plausible scenario of weak to moderate selection there exist two alternate pathways for compensatory substitution. In one pathway, a deleterious mutation becomes fixed prior to occurrence of the compensatory mutation. In the other, the two loci are simultaneously polymorphic. The rates of compensatory substitution along these two pathways and their relative probabilities are functions of the population size, selection strength, mutation rate, and recombination rate. In this paper these rates and path probabilities are derived analytically and verified using population genetic simulations. The expected time durations of these two paths are similar when selection is moderate, but not when selection is weak. The effect of recombination on the dynamics of the substitution process are explored using simulation. Using the derived rates, a phylogenetic substitution model of the compensatory evolution process is presented that could be used for inference of population genetic parameters from interspecific data.