TcrDesign: De novo design of epitope specific full-length T cell receptors

ABSTRACT T cell receptors (TCRs) are essential for adaptive immune recognition. Recently computational tools have been developed to predict the interactions between TCR and epitope, and artificial intelligence models have been proposed to generate the complementarity determining region 3 (CDR3) region of β chain TCR. However, de novo design of experimentally validated, functional, full-length epitope-specific TCRs remains a significant challenge. Here we developed TcrDesign, a deep learning framework based on large-scale unlabeled TCR and epitope datasets to generate epitope-specific, full-length TCRs. TcrDesign comprises two modules: TcrDesign-B for TCR-pMHC binding prediction with state-of-the-art accuracy, and TcrDesign-G for functional full-length TCR sequence generation. Pre-trained on large-scale unlabeled datasets using transformer-based architectures, TcrDesign achieves state-of-the-art performance in both TCR-epitope binding prediction and de novo TCR sequence generation. Furthermore, epitope-major histocompatibility complex (MHC) binding and functional activation of TcrDesign-generated TCRs were experimentally validated. TcrDesign provides an efficient and modular approach for designing epitope-specific full-length TCRs, with experimental validation confirming its utility. Competing Interest Statement The authors have declared no competing interest.