#30 · Metabolomics

ML-Guided Proteomics Methods Resolve Site-Level Structural and PTM Changes

The common thread is computational inference of molecular structure, modification state, and mechanism from proteomic measurements. A benchmark of peptide-to-protein aggregation strategies for site-level PTM and LiP-MS quantification introduces a site-centric roll-up for proteolytic fragments. Separately, diazirine photo-cross-links are shown to be MS-cleavable, and a side-chain fingerprint ML model with a rule-based filter improves XL-MS site identification and FDR control. A Cell Painting and thermal proteome profiling study integrates morphological profiles with proteome-wide thermal data to build PPI networks for drug target and mechanism-of-action inference. A preprint applies AI-guided screening to 2,235 protein fragments, with 18 tested experimentally, to design regulators of biomolecular condensates. What is new across these items is the use of ML and integrative data to move from detection toward residue-level and network-level interpretation.