Analysis of selection in protein-coding sequences accounting for common biases

Abstract

The evolution of protein-coding genes is usually driven by selective processes, which favor some evolutionary trajectories over others, optimizing the subsequent protein stability and activity. The analysis of selection in this type of genetic data is broadly performed with the metric nonsynonymous/synonymous substitution rate ratio (dN/dS). However, most of the well-established methodologies to estimate this metric make crucial assumptions, such as lack of recombination or invariable codon frequencies along genes, which can bias the estimation. Here, we review the most relevant biases in the dN/dS estimation and provide a detailed guide to estimate this metric using state-of-the-art procedures that account for such biases, along with illustrative practical examples and recommendations. We also discuss the traditional interpretation of the estimated dN/dS emphasizing the importance of considering complementary biological information such as the role of the observed substitutions on the stability and function of proteins. This review is oriented to help evolutionary biologists that aim to accurately estimate selection in protein-coding sequences.

ProteinEvolverABC: coestimation of recombination and substitution rates in protein sequences by approximate Bayesian computation

“…A recent example is the problematic detection of recombination along the SARS-CoV-2 genome due to the relatively low levels of genetic diversity displayed by this virus ( VanInsberghe et al , 2021 ). The estimation of the recombination rate can be affected by the substitution rate and vice versa [note that the recombination rate can exceed the substitution rate in some organisms such as HIV ( Perez-Losada et al , 2015 ; Shriner et al , 2004 ) and affect estimations of diverse evolutionary parameters such as the synonymous and non-synonymous substitution rates ( Anisimova et al , 2003 ; Arenas and Posada, 2010 ; Del Amparo et al , 2021 ; Shriner et al , 2003 )]. Therefore, the coestimation of both parameters is convenient due to accounting for their parameters interactions ( Lopes et al , 2014 ).…”

“…In such situations, an alternative is the approximate Bayesian computation (ABC) approach ( Beaumont, 2010 ; Csillery et al , 2010 ), which considers computer simulations under complex evolutionary scenarios followed by a statistical adjustment for the estimation of parameters without the need of a likelihood function. Hence, taking advantage of ABC, we previously developed an ABC framework for the joint estimation of recombination, selection [by the non-synonymous/synonymous rate ratio, dN/dS ( Del Amparo et al , 2021 )] and substitution rates in coding sequences that outperformed other methods (including maximum-likelihood methods) in the estimation of the recombination rate ( Arenas et al , 2015 ; Lopes et al , 2014 ). Here, I adapted that ABC method to the evolutionary analysis of protein sequences and I implemented it in a user-friendly computational framework called ProteinEvolverABC .…”

Phylogeography of Tridentiger bifasciatus (Gobiidae) in the Northwestern Pacific

“…The values of NSS/SS in lineage II were lower than that in lineage I, excluding one specimen in population YT. Thus, our study suggests that the two mtDNA divergent lineages might be maintained due to purifying selection (Ka/Ks < 1; Amparo et al, 2021) and different strengths of selection in the two geographic regions.…”

Resolving Complexities in Taxonomic Lineages of the Organellar and Nuclear Genomes of Galdieria through Comparative Phylogenomic Analysis

“…Analyzing the rate of the substitutions occurring in a protein can identify information about which selective pressures are happening [87]. The calculation of dN/dS can therefore help to identify genes that are under particular biochemical or ecological constraints, or conversely putative proteins involved in survival adaptation.…”