TCR-ESM

TCR-ESM predicts interactions between T-cell receptors (TCRs), peptides, and major histocompatibility complexes (pMHC) using protein language embeddings to identify cognate targets relevant to T-cell therapy development.

Key Features:

• Protein Language Embeddings: Utilizes peptide embeddings learned from the Evolutionary Scale Modeling (ESM) large-scale protein language model to capture evolutionary-scale biological information for TCR-pMHC binding prediction.
• Complementarity-Determining Region 3 (CDR3): Emphasizes the hypervariable CDR3 regions and incorporates both CDR3α and CDR3β chain sequence information to improve peptide recognition specificity.
• Dual Chain Contribution: Accounts for contributions from both TCR chains to binding specificity, with relative importance varying depending on the specific peptide-MHC target.
• Importance of MHC Information: Demonstrates the critical importance of MHC information across datasets for accurate modeling of TCR-peptide binding.
• Performance and Generalizability: Outperforms existing predictors on external datasets, indicating robust generalizability across diverse immunological data.

Scientific Applications:

• T-cell Therapy Target Identification: Predicts cognate TCR-pMHC interactions to aid identification of potential targets for T-cell therapies.
• TCR Specificity Analysis: Facilitates analysis of factors driving TCR specificity by integrating sequence and MHC information.
• Personalized Medicine and Immunological Research: Supports personalized medicine approaches and research requiring accurate TCR targeting and cross-dataset generalization.

Methodology:

Uses protein language embeddings from the Evolutionary Scale Modeling (ESM) protein language model, applying peptide embeddings and incorporating CDR3α and CDR3β chain sequence information.