Development of a deep learning model for predicting the interaction between T cell receptors and epitopes

• Grant number: 24/20196-1
• Support Opportunities: Scholarships in Brazil - Master
• Start date: July 01, 2025
• End date: December 31, 2026
• Field of knowledge: Biological Sciences - Biochemistry - Chemistry of Macromolecules
• Principal Investigator: Helder Veras Ribeiro Filho
• Grantee: Samuel Chagas de Assis
• Host Institution: Centro Nacional de Pesquisa em Energia e Materiais (CNPEM). Ministério da Ciência, Tecnologia e Inovação (Brasil). Campinas , SP, Brazil

Abstract

T lymphocytes, through T-cell receptors (TCR), play a fundamental role in the immune response by recognizing specific antigenic epitopes in the form of peptides presented by the major histocompatibility complex (pMHC). Understanding the interaction between TCR and pMHC allows for insights into TCR specificity for antigens, and consequently, the determination of the function of TCR repertoires in the body, which is useful in diagnosing and understanding the immune response against diseases. Furthermore, the ability to determine TCR recognition of antigens enables the design or prioritization of new specific TCRs that can be used in T cell-based immunotherapies, as well as a more refined assessment of antigenic epitopes for T cell-based vaccines, with the aim of reducing cross-reactivity. Therefore, in this project, we will develop a computational tool based on deep learning, using graph neural networks to predict the interaction between TCR and epitopes presented by MHC. To overcome the challenges faced by existing predictors, which limit their accuracy and generalization capability for unseen epitopes, our predictor will integrate 3D structural information of TCR and pMHC and use existing implementations of graph networks with attention mechanisms, as well as pre-trained protein language models, to provide enriched and contextualized representations of the amino acids in these proteins. By using all relevant regions of the TCR and pMHC for this interaction, we expect the model to learn relationships between amino acids that are determinants for the interaction.