FURI | Fall 2025

Fine-tuning RFdiffusion for Structure-based Design of Epitope-specific T Cell Receptors

T cell receptors (TCRs) recognize specific molecular fragments, called epitopes, that play a critical role in immune recognition and therapy design. However, designing novel TCRs that bind to new epitopes remains challenging due to limited structural data and sequence diversity. This research fine-tunes RoseTTAFold Diffusion (RFdiffusion), a deep learning model for 3D protein generation, to create epitope-conditioned TCR backbones. By coupling these structures with ProteinMPNN for sequence optimization, the approach enables de novo design of high-affinity TCRs, advancing computational immunotherapy and personalized medicine.