#40 · ML / AI methods

Computational methods resolve protein and molecular variants beyond canonical sequence…

The common thread is computational resolution of protein and molecular variation that canonical sequence-to-form annotation does not capture. ProteoformDB generates a human database cataloging over one million unique single-site proteoforms for proteomics. VInSMoC applies scalable mass spectral database search to identify molecular variants with significance estimation, applied to 483M GNPS spectra. Morpheus predicts fold-switching behaviour from fragments across proteomes. A multitiered pipeline integrating sequence bioinformatics, biophysics, and AI/peptide structure prediction scores and characterizes 85 miniproteins. What is new is systematic scoring of variant and conformational states at proteome and spectral-library scale, extending from proteoform catalogs to fold-switch and miniprotein characterization. The items are heterogeneous; the shared element is extension of function-relevant analysis beyond canonical assumptions.