Limitations of de novo peptide sequencing in resolving sequence ambiguity.

Eight state-of-the-art de novo peptide sequencing models were benchmarked across three proteomics datasets, showing variable performance but substantial overlap in correct identifications.

• Why it matters: Ambiguous spectra from incomplete fragment ion coverage can prevent unambiguous peptide identification, and it is unclear which deep learning models best handle this ambiguity.
• What they did: The study benchmarked eight models on three public proteomics datasets, compared them with established metrics, measured inter-model agreement, and tested post-processing via iterative refinement, rescoring, and reranking.
• The result: Post-processing produced no or only modest gains, and most sequencing errors were model-independent, linked to limited fragment ion coverage, mirroring issues seen in database searches with large search spaces.

AI summary of the abstract · may contain errors · not medical advice · read the paper.