Modeling TCR-Epitope Recognition Specificity: What We Should Learn to Succeed

T-cell recognition of infected or malignant cells is central to both spontaneous and therapy-induced cellular immune responses against pathogens and cancer. This recognition is elicited by the interaction between T-Cell Receptors (TCRs) and epitopes, which consist of antigenic peptides displayed on major histocompatibility complex molecules. TCR-epitope interactions are characterized by high diversity in TCR and epitope sequences and high structural flexibility in TCR loops. As a result, deciphering the rules of TCR-epitope recognition specificity and accurately predicting these interactions remains challenging. Here, we review the different strategies developed to predict TCR-epitope recognition, classify the principal computational frameworks, examine the data modalities on which they depend and discuss their current limitations. We then synthesize key conceptual insights that have emerged from recent research and outline how these lessons should inform the design of future experiments and next-generation computational tools.