DEX: a consensus-based amino acid exchangeability measure for improved codon substitution modelling.

Area of Science:

• Molecular Evolution
• Genomics
• Bioinformatics

Background:

• Amino acid similarity is crucial for understanding molecular evolution, yet existing distance measures are underutilized due to a lack of consensus on accuracy.
• Physicochemically similar amino acids exhibit more frequent substitutions, but quantifying these relationships for evolutionary modeling remains challenging.

Purpose of the Study:

• To systematically compare the performance of 30 amino acid distance measures in molecular evolution models.
• To introduce a novel, improved amino acid distance measure based on recent deep mutational scanning data.
• To evaluate the utility of variant effect predictors versus amino acid distance measures in predicting substitution frequencies.

Main Methods:

• Assessed 30 amino acid distance measures, including a new deep mutational scanning-derived measure, against codon substitution models.
• Utilized alignments from diverse lineages (Streptococcus, Drosophila, mammals) and variant data (E. coli, human).
• Employed the DISTATIS approach to construct and test consensus measures from top-performing individual measures.

Main Results:

• Experimentally-derived amino acid distance measures, particularly the new measure and the existing experimental exchangeability (EX) measure, demonstrated the best fit to codon substitution patterns.
• A consensus measure, named DEX, combining the new measure and EX, performed optimally across diverse datasets.
• Site-specific variant effect predictors were less effective than amino acid distance measures for predicting mean substitution frequencies but informative for identifying highly deleterious mutations.

Conclusions:

• The DEX measure offers an improved, general-purpose tool for molecular evolution models, outperforming existing measures in fitting substitution patterns.
• Experimentally-derived measures are superior for capturing evolutionary substitution dynamics compared to traditional physicochemical measures.
• While variant effect predictors have distinct applications, amino acid distance measures remain essential for large-scale evolutionary analyses.