Abstract
We present a method that quantifies nearly all pairwise amino acid substitutions, revealing the rates and spectra of mistranslation per amino acid and per codon. While determining error rates is illuminating, understanding why certain mistranslation events are enriched requires further exploration, particularly in relation to the translation-accuracy hypothesis. We found that codons favored in highly expressed genes are not predictive of translational fidelity and are not selected for translational accuracy. We evaluated our method using historically characterized ribosomal variants of Escherichia coli reported to differ in translation fidelity. However, we found no significant differences among their overall translation-error rates, estimated here to be ~2 per 1,000 amino acids. Instead they each exhibited unique mistranslation profiles, which likely led to prior overestimates of the translation error-rate in the putative error-prone variant. Together, these results underscore the significance of a direct, proteome-wide measure of translation fidelity.