pNovo 3

pNovo 3 performs de novo peptide sequencing from tandem mass spectrometry (MS/MS) data using a learning-to-rank framework and pDeep-predicted theoretical spectra to improve peptide identification precision.

Key Features:

• Learning-to-Rank Framework: Uses a learning-to-rank approach to distinguish and rank similar peptide candidates for each MS/MS spectrum, addressing low ion coverage challenges.
• Deep Learning Integration (pDeep): Utilizes the pDeep algorithm to predict theoretical spectra via deep learning for similarity comparison with experimental spectra.
• Three Similarity Metrics: Employs three explicit metrics to measure similarity between experimental and theoretical spectra.
• Database-Independent Sequencing: Performs de novo sequencing and enables assembly/identification of unknown proteins without relying on pre-existing databases.
• Improved Recall and Precision: Reported recall improvements of 29–102% and precision increases of 11–89% on benchmark datasets from six species.
• Superior Performance over DeepNovo: Identified 21–50% more spectra than DeepNovo across nine datasets.

Scientific Applications:

• De novo peptide sequencing: Determines peptide sequences directly from MS/MS data without database dependence.
• Unknown protein discovery: Supports identification and assembly of proteins not present in existing sequence databases.
• Cross-species proteomics: Facilitates exploration of proteomic diversity across species through improved peptide identification.

Methodology:

Applies a learning-to-rank algorithm to rank candidate peptides per MS/MS spectrum, compares experimental spectra to pDeep-predicted theoretical spectra using three similarity metrics.