Prosit

Prosit predicts peptide tandem mass spectra, chromatographic retention times, and fragment ion intensities using a deep neural network to improve peptide identification and spectral-library generation for mass-spectrometry-based proteomics.

Key Features:

• Deep Neural Network Architecture: Trained on a dataset of 550,000 tryptic peptides and over 21 million high-quality tandem mass spectra to predict chromatographic retention times and fragment ion intensities.
• Enhanced Predictive Accuracy: Produces predictions that surpass experimental data quality, improving peptide identification and quantification.
• Integration with Database Search Pipelines: Can be incorporated into database search pipelines to increase identification rates while maintaining false discovery rates at more than ten times lower levels.
• Versatility Across Proteases: Accurately predicts spectra for proteases beyond trypsin, supporting diverse digestion strategies.
• Support for Data-Independent Acquisition: Generates spectral libraries tailored for data-independent acquisition (DIA) workflows, aiding analysis of complex samples including metaproteomes.
• Integration with ProteomicsDB: Enables re-scoring of search results and generation of custom spectral libraries based solely on peptide sequences within ProteomicsDB.

Scientific Applications:

• Peptide Identification and Quantification: Improves identification rates and quantitative reliability in mass-spectrometry-based proteomics while reducing false discovery rates.
• Spectral Library Generation for DIA: Produces spectral libraries for data-independent acquisition to support analysis of complex and metaproteomic samples.
• Cross-Organism and Proteome-Wide Studies: Supports comprehensive proteomic analyses across diverse organisms via integration with ProteomicsDB and sequence-based library generation.
• Protein Function, Interaction, and Dynamics Studies: Enhances downstream analyses of protein functions, interactions, and dynamics by providing high-quality predicted spectra and retention times.

Methodology:

Uses a deep neural network trained on 550,000 tryptic peptides and over 21 million high-quality tandem mass spectra to predict chromatographic retention times and fragment ion intensities, generates spectral libraries, and enables re-scoring of database search results and integration into search pipelines.