#26 · Proteomics

ML rescoring and mass offset search refine peptide and PTM identification from MS/MS

The dominant thread is the use of learned or model-based components on tandem mass spectra to improve peptide identification and interpretation. quantms-rescoring, a preprint, adds ML-based rescoring with fine-tuned MS2 and retention-time predictors on top of multi-engine search using SAGE, COMET, and MSGF+ within a cloud-native workflow, spanning protein quantification, immunopeptidomics, and PTM experiments. The MSFragger detailed mass offset search is rule-based rather than learned, adding per-offset site restrictions and fragmentation rules for complex PTM interpretation. The Tau cross-linking framework is methodologically adjacent: a network-based approach to XL-MS data that infers structural differences between disordered Tau and phosphorylated Tau. What is new is the pairing of learned spectral scoring with multi-engine search and rescoring, a direction that currently rests on a single preprint.