nHLAPred

nHLAPred predicts MHC class I–restricted T cell epitopes by combining quantitative matrix (QM) and artificial neural network (ANN) methods to identify binding peptides across 67 MHC class I alleles and potential C-terminal proteasomal cleavage sites.

Key Features:

• Extensive Allele Coverage: Supports predictions for 67 MHC class I alleles.
• Hybrid Prediction Methodology: Employs a QM-based method for 47 alleles with ≥15 known binders and an ANN-based approach for 30 alleles with ≥40 binders.
• Accuracy Improvement: The integrated QM+ANN hybrid increases prediction accuracy by 6% over individual methods and achieves an average accuracy of 92.8%.
• Proteasomal Cleavage Site Prediction: Identifies proteasomal cleavage sites within antigen sequences, focusing on peptides with C-terminal cleavage sites as potential T cell epitopes.
• Performance Evaluation: Validated using jack-knife tests and independent datasets, demonstrating superior performance relative to existing MHC binder prediction methods.
• Promiscuous Binder Detection: Identifies promiscuous MHC binding regions from antigen sequences.

Scientific Applications:

• Immunotherapy Development: Aids design of personalized vaccines and immunotherapies by predicting potential T cell epitopes across diverse MHC alleles.
• Vaccine Design: Predicts promiscuous binders to enhance vaccine coverage across heterogeneous populations.
• Disease Research: Supports studies of autoimmune diseases, infectious diseases, and cancer immunology by characterizing peptide–MHC interactions.

Methodology:

Combines QM-based and ANN-based prediction methods (QM applied to 47 alleles with ≥15 known binders; ANN applied to 30 alleles with ≥40 binders) and uses jack-knife tests and independent dataset validation.