PSM Rescoring with MS²Rescore

Download a real HeLa LC-MS/MS dataset and rescore MaxQuant PSMs using MS2PIP fragment ion predictions, DeepLC retention time predictions, and mokapot — gaining +16 % identifications at 1 % FDR.

Database search engines assign each spectrum a score based on mass accuracy and matched fragment ions, but leave orthogonal information unused. MS²Rescore adds features from predictive models — fragment ion intensities (MS2PIP), retention time (DeepLC), and basic search-engine metadata — and re-ranks PSMs with a semi-supervised SVM (mokapot).

This tutorial downloads a real 7-minute HeLa gradient run (MaxQuant search, ~111 MB of input data), runs the full MS²Rescore pipeline, and shows the improvement in identifications at 1 % FDR.

The source notebook is available in the CompOmics website repository.