Big data analysis of mitochondrial DNA substitution models: A regression approach elucidating the effects of codon position and neighboring nucleotides

Abstract

AbstractWe build on the up-to-date version of Phylotree, a comprehensive and continuously updating phylogeny of global human mtDNA variations (van Oven and Kayser 2009), to better understand the substitution mechanism of the mitochondrial DNA (mtDNA) and its most influential factors. We do so by composing Poisson and negative-binomial regression models relating the rate of occurrence of mtDNA substitutions to various factors. Important factors we identify include the identity of the codon at each position, confirming previous findings about the biological significance of different codons for the same amino acid. Importantly, we also identify a significant effect of neighboring sites. This effect cannot be attributed solely to CpG pairs. A similar effect of neighboring sites was recently described for autosomal DNA substitutions, and we speculate it is related to the basic mutational mechanism itself. Once codon composition and context are taken into account, there is no significant difference in substitution rate between different genes in mtDNA.