Nokoi

Nokoi performs decoy-free binary classification of peptide-to-spectrum matches (PSMs) to validate peptide identifications in shotgun proteomics, proteogenomics, and metaproteomics.

Key Features:

• Decoy-Free Approach: Eliminates the need for decoy sequences to separate correct and incorrect PSMs.
• Binary Classification: Uses a machine learning binary classifier to assign PSMs as correct or incorrect.
• Mascot Rank Labels: Trains using ranks provided by the Mascot search engine as labels for PSMs.
• Heterogeneous Training Data: Trained on an extensive and heterogeneous dataset to improve robustness across conditions.
• Performance Comparison: Demonstrates higher accuracy than Mascot and comparable accuracy to Percolator while providing significantly higher processing speeds.

Scientific Applications:

• Large-scale proteomics: Accelerates and validates peptide identification in high-throughput proteomics datasets.
• Proteogenomics: Validates PSMs in proteogenomic analyses without relying on decoy databases.
• Metaproteomics: Provides PSM validation for complex community-derived metaproteomic samples without decoy-based approaches.
• PSM validation: Supplies an alternative method for distinguishing correct and incorrect peptide identifications.

Methodology:

Nokoi implements a machine learning–based binary classifier trained on extensive heterogeneous datasets using ranks from the Mascot search engine as training labels.