NeuRiPP

NeuRiPP classifies peptide sequences as putative precursor peptides (PPs) for ribosomally synthesized and post-translationally modified peptides (RiPPs) to enable identification of PPs within biosynthetic gene clusters (BGCs).

Key Features:

• Neural Network Architecture: Employs neural network architectures tailored for peptide classification trained on PP sequence data.
• Comprehensive Training Data: Trained on datasets comprising high-confidence putative PP sequences and experimentally verified PPs from established RiPP databases.
• Independence from Genetic Context: Classifies candidate PPs without requiring context-dependent information about the surrounding genetic environment.
• Enhanced Identification Capability: Identifies PP sequences in more putative RiPP clusters compared to existing tools while maintaining the same Hidden Markov Model (HMM) hit accuracy.
• Utility for Novel RiPP Classes: Demonstrates improved detection of precursor peptides from recently characterized novel RiPP classes.

Scientific Applications:

• RiPP precursor peptide identification: Classifies peptide sequences as putative PPs to support annotation of RiPP biosynthetic gene clusters.
• BGC mining: Enables screening of RiPP BGC datasets for candidate precursor peptides independent of genetic context.
• Novel RiPP class discovery: Facilitates detection and characterization of precursor peptides from newly described RiPP classes.
• Natural product discovery: Supports prioritization of candidate RiPPs for downstream experimental characterization and bioactive compound discovery.

Methodology:

NeuRiPP trains neural network architectures on datasets of high-confidence putative and experimentally verified PP sequences from established RiPP databases, evaluates classification performance on RiPP BGC datasets, and reports identification rates while comparing Hidden Markov Model (HMM) hit accuracy; the classifier operates independently of genetic context.