bioRxiv2023
Amino acid substitutions (AAS) change a protein from its genome-expected sequence. Accumulation of substitutions in proteins underlie numerous diseases and antibiotic mechanisms. Accurate global detection of substitutions and their frequencies would help characterize these mechanisms. Measurement of AAS using shotgun proteomics is attractive due to its high sensitivity and untargeted acquisition. However, identifying substituted peptide-spectra requires search strategies that extrapolate beyond the genome, which can introduce bias. To characterize this bias, we constructed a "ground-truth" approach using the similarities between the Escherichia coli and Salmonella typhimurium proteomes to effectively model the complexity of distinguishing substitutions from genomic peptides. Shotgun proteomics on combined whole cell lysates from both organisms generated a library representing nearly 100,000 peptide-spectra and 4,161 distinct peptide sequences corresponding to genome-level single AAS with defined stoichiometry. We tested the ability to identify S. typhimurium peptide-spectra using only the E. coli genome in substitution-tolerant database searching. Overall, 64.1% of library peptides were correctly identified. We observed a wide range of identification efficiencies based on the specific AAS, but no inherent bias from stoichiometry of the substitution. Short peptides and substitutions near peptide termini, which require specific diagnostic ions for unambiguous identification, are matched with below-average frequency. We also identified "scissor substitutions" that gain or lose protease cleavage sites. Although scissor substitutions are chemically distinct from the genomic peptide, they had poor identification efficiency. This ground-truth AAS library identifies multiple sources of bias in AAS peptide-spectra identification and sets expectations for the application of shotgun proteomics to testing AAS hypotheses. Significance statementHigh-fidelity decoding of the genome is essential for life. Mistranslation leads to amino acid substitutions, which can disrupt protein folding and function, and impact cell fitness. Detection of mistranslated protein products necessitates robust and non-biased approaches. Proteomics is a promising solution, but identifying non-genomic peptide-spectra is a severe bioinformatics challenge. We created a ground-truth library of substituted amino acid peptides by mixing two closely related bacteria in a single sample. We quantitatively defined the degree to which informatics could correctly distinguish substituted peptides when single-organism databases are present. This approach defines intrinsic and informatics limits in substitution detection in shotgun proteomics and identifies previously overlooked challenges with identifying "scissor substitutions".