Inferring the selective history of CNVs using a maximum likelihood model.

Abstract

Copy number variations (CNVs) - structural variations generated by deletion and/or duplication that result in a change in DNA dosage - are prevalent in nature. CNVs can drastically affect the phenotype of an organism and have been shown to be both involved in genetic disorders and be used as raw material in adaptive evolution. Unlike single-nucleotide variations, the often large and varied effects of CNVs on phenotype hinders our ability to infer their selective advantage based on the population genetics data. Here, we present a likelihood-based approach, dubbed PoMoCNV (POlymorphism-aware phylogenetic MOdel for CNVs), that estimates the evolutionary parameters such as mutation rates among different copy numbers and relative fitness loss per copy deletion at a genomic locus based on population genetics data. As a case study, we analyze the genomics data of 40 strains of Caenorhabditis elegans, representing four different populations. We take advantage of the data on chromatin accessibility to interpret the mutation rate and fitness of copy numbers, as inferred by PoMoCNV, specifically in open or closed chromatin loci. We further test the reliability of PoMoCNV by estimating the evolutionary parameters of CNVs for mutation-accumulation experiments in C. elegans with varying levels of genetic drift.