DeepRescore

"DeepRescore" is a post-processing tool to enhance the sensitivity and reliability of peptide identification in mass spectrometry (MS)-based immunopeptideomics, specifically for the identification of major histocompatibility complex (MHC)-binding peptides. Immunopeptidomics experiments, pivotal for understanding immune responses, often face challenges due to the lack of enzymatic digestion at specific residues, leading to inflated search space, high false positive rates, and low sensitivity in peptide identification. DeepRescore addresses these challenges by incorporating advanced deep-learning predictions into the peptide identification process.

Key Features:

- Deep Learning Predictions: DeepRescore integrates peptide features derived from deep learning, including accurate predictions of retention time and MS/MS spectra, alongside previously utilized features. This integration significantly improves the tool's ability to accurately identify MHC-binding peptides and neoantigens.

- Enhanced Sensitivity and Reliability: By rescoring peptide-spectrum matches (PSMs) with these advanced features, DeepRescore substantially increases the sensitivity and reliability of peptide identifications. This enhancement is crucial for accurately detecting potential immune targets and biomarkers.

- Performance Validation: The effectiveness of DeepRescore has been validated using two public immunopeptidomics datasets. The results demonstrate that rescoring with DeepRescore outperforms existing methods, largely due to including deep learning-derived features.

- Accessibility and Usability: Developed using NextFlow and Docker, DeepRescore is designed for ease of use and accessibility.