iTCep

iTCep predicts T cell epitopes using a deep learning framework that models peptide–T-cell receptor (TCR) interactions and TCR residue preferences to improve neoantigen identification.

Key Features:

• Deep learning framework: Uses a deep learning framework to predict peptide–TCR interactions for epitope identification.
• Novel encoding: Employs a novel encoding approach that captures intricate interactions between peptides and T-cell receptors and residue preferences within TCRs.
• Feature-level fusion strategy: Derives fusion features via a feature-level fusion strategy to combine peptide and TCR information.
• Prediction modes: Supports predictions for peptide–TCR pairs and for peptides alone.
• Binding affinity consideration: Considers specific binding affinities within TCRs when evaluating peptide-TCR interactions.
• Performance metrics: Achieves area under the curve (AUC) up to 0.96 on testing datasets and maintains above 0.86 across independent datasets.
• Improved neoantigen identification: Enhances predictive performance for identifying epitopes that can trigger immune responses.

Scientific Applications:

• Personalized cancer immunotherapy: Supports selection of candidate neoantigens for personalized cancer immunotherapy strategies.
• Neoantigen identification: Improves identification of neoantigens through modeling of peptide–TCR interactions.
• TCR binding specificity prediction: Predicts T-cell receptor binding specificities for given peptides.
• Epitope selection: Identifies epitopes with higher likelihood to trigger immune responses.

Methodology:

Applies a deep learning framework with a novel encoding of peptide–T-cell receptor interactions and a feature-level fusion strategy to derive fusion features for prediction of peptide–TCR interactions, supporting both peptide–TCR pair and peptide-only prediction modes.