T cell receptor cross-reactivity prediction improved by a comprehensive mutational scan database.

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Mapping T cell receptor (TCR) targets is essential for predicting pathogenic escape and off-target effects in TCR therapies.
• Current challenges include a lack of unbiased benchmarking datasets and sensitive computational methods for small-peptide mutations.

Purpose of the Study:

• To create a comprehensive database and develop a computational model for accurate TCR target mapping.
• To enable efficient identification of TCR targets for novel TCRs.

Main Methods:

• Curated the Benchmark for Activation of T cells with Cross-reactive Avidity for Epitopes (BATCAVE) database, containing over 22,000 TCR-peptide pairs from single-amino-acid mutational assays.
• Developed Bayesian Inference of Activation of TCR by Mutant Antigens (BATMAN), an interpretable Bayesian model trained on BATCAVE.
• Implemented an active learning extension for efficient target mapping of novel TCRs.

Main Results:

• BATMAN outperforms existing methods in predicting TCR-peptide interactions.
• The model reveals key structural and biochemical predictors of TCR-peptide binding.
• BATMAN accurately predicts polyclonal T cell responses and identifies TCR targets with high sequence dissimilarity.

Conclusions:

• The BATCAVE database and BATMAN model provide a robust framework for TCR target mapping.
• This approach enhances the prediction of TCR-peptide interactions and facilitates the development of safer and more effective TCR therapies.