NetCleave

NetCleave predicts C-terminal antigen processing to identify peptide C-termini that determine subsequent presentation by MHC-I and MHC-II pathways.

Key Features:

• Algorithm Architecture: Neural network trained on 46 distinct physicochemical descriptors of cleavage-site amino acids.
• QSAR Descriptors: Encodes amino acids using Quantitative Structure-Activity Relationship descriptors capturing steric, electrostatic, and hydrophobic properties.
• Predictive Performance: Reported AUC of 0.91 for MHC-I predictions and 0.66 for MHC-II predictions.
• Retrainable: Supports retraining the model with new datasets.
• Database Integration: Maps peptides reported in the IEDB to protein sequences in UniProt/UniParc.

Scientific Applications:

• Immunogenic peptide discovery: Identifies candidate antigenic peptides by predicting C-terminal processing sites relevant for MHC presentation.
• Vaccine and therapeutic design: Prioritizes peptides for vaccine and therapeutic antigen development based on predicted processing.
• Evaluation of immunogenicity datasets: Used to evaluate independent datasets including H2-Db, H2-Kb, HLA-A*02:01, and HLA-B:07:02.
• Immunotherapy and personalized medicine: Supports peptide prioritization for downstream immunotherapy and personalized antigen-presentation analyses.

Methodology:

Maps IEDB peptides to UniProt/UniParc protein sequences, identifies C-terminal cleavage sites, encodes amino acid sequences with QSAR descriptors, and uses a neural network trained on 46 physicochemical descriptors to predict C-terminal processing.