AI-driven computational methods and benchmarking for T-cell antigen identification - PubMed
5 hours ago
- #AI in immunology
- #mRNA vaccines
- #T-cell antigen identification
- The rise of mRNA vaccines underscores the importance of T-cell antigen identification in vaccinology and personalized medicine.
- T-cell recognition involves a ternary interaction between TCR, MHC molecule, and peptide antigen (pMHC complex).
- AI-driven computational methods are crucial for predicting TCR-pMHC binding.
- The review categorizes methods for MHC-I, MHC-II, and TCR-pMHC binding prediction and evaluates foundational data resources.
- A benchmarking study of 18 TCR-pMHC prediction models reveals a significant generalization gap under out-of-distribution (OOD) conditions.
- Current models show marginal predictive gains when faced with novel epitope variants.
- The study highlights the need for improved structural modeling, multi-omics data integration, and generative models for de novo TCR design.
- Advancing these computational methods could transition immunoinformatics from prediction to rational design.