MS2Rescore

MS2Rescore improves peptide-spectrum match (PSM) rescoring in mass spectrometry-based proteomics to enhance identification of MHC-presented peptides in immunopeptidomics and to address challenges in nontryptic peptide identification.

Key Features:

• Model integration: Integrates MS²PIP, DeepLC, and Percolator to combine peak intensity prediction, retention time prediction, and machine-learning rescoring for PSMs.
• Retrained MS²PIP models: Uses retrained MS²PIP models specifically adapted to predict fragment ion peak intensities for nontryptic peptides.
• Improved predictive accuracy: Enhances prediction accuracy for immunopeptides and tryptic peptides, leading to more reliable PSM evaluation.
• Enhanced identification rates: Produces a 46% increase in spectrum identification rate and a 36% increase in unique peptide identifications compared to standard Percolator rescoring at a 1% FDR.
• Outperformance of current approaches: Surpasses existing state-of-the-art methods tailored for immunopeptide identification.

Scientific Applications:

• Immunopeptidomics: Facilitates identification of MHC-presented peptides across cell types to support epitope discovery for applications such as anti-cancer vaccine research.
• Proteomics research: Applies to broader proteomics datasets to improve peptide identification accuracy for both tryptic and nontryptic peptides.

Methodology:

Integrates MS²PIP, DeepLC, and Percolator and uses retrained MS²PIP models to predict fragment ion peak intensities for nontryptic peptides as part of PSM rescoring.