Structure-based prediction of T cell receptor:peptide-MHC interactions

Area of Science:

• Immunology
• Structural Biology
• Computational Biology

Background:

• T cell receptor (TCR) and peptide-MHC interactions are central to adaptive immunity.
• Predicting TCR:peptide-MHC specificity is challenging due to diverse recognition modes and limited data.

Purpose of the Study:

• To evaluate deep learning-based structural modeling for predicting TCR epitope specificity.
• To assess the accuracy of AlphaFold in modeling TCR:peptide-MHC interactions for epitope discrimination.

Main Methods:

• Utilized a specialized version of the AlphaFold neural network predictor.
• Generated structural models of TCR:peptide-MHC complexes.
• Assessed the ability of models to discriminate correct from incorrect peptide epitopes.

Main Results:

• AlphaFold successfully generated TCR:peptide-MHC interaction models.
• These models demonstrated substantial accuracy in discriminating correct peptide epitopes.
• Deep learning structural modeling shows potential for predicting TCR specificity.

Conclusions:

• Deep learning-based structural modeling offers a promising approach for predicting TCR:peptide-MHC specificity.
• Further development is needed for widespread practical application of these predictive models.
• This methodology could advance understanding of adaptive immune responses.