Background: Immunotherapy has transformed cancer care despite heterogeneity in clinical responses. T cells' recognition of tumor antigens is central to the effectiveness of cancer immunotherapies. But little remains understood about the tumor antigens that are immunogenic and the T cells that recognize them. Accordingly, studying TCR-antigen interactions is the key to understanding the biology of tumor immune responses and reinvigorating immunotherapies. However, traditional experimental investigation of the interaction between tumor antigens and T cells is labor intensive, costly, time-consuming, and presents excessively low validation rates of ~1%. Therefore, it is critical to develop mach...

Read More

Background: Immunotherapy has transformed cancer care despite heterogeneity in clinical responses. T cells' recognition of tumor antigens is central to the effectiveness of cancer immunotherapies. But little remains understood about the tumor antigens that are immunogenic and the T cells that recognize them. Accordingly, studying TCR-antigen interactions is the key to understanding the biology of tumor immune responses and reinvigorating immunotherapies. However, traditional experimental investigation of the interaction between tumor antigens and T cells is labor intensive, costly, time-consuming, and presents excessively low validation rates of ~1%. Therefore, it is critical to develop machine learning approaches that can predict the binding between TCRs and T cell antigens in an accurate and cost-effective manner. Overall goal: We have established a multi-disciplinary team of diverse expertise, including Dr. Tao Wang (UTSW), Dr. Junzhou Huang (UTA) and Dr. Alexandre Reuben (MDACC). We will develop a suite of knowledge-guided deep learning models to predict the binding between T cell receptors (TCRs) and T cell antigens. Aim1: Build a knowledge-guided TCR structure prediction model to obtain TCR structural embeddings. Aim2: Build a knowledge-guided TCR-pMHC binding prediction workflow based on TCR and pMHC sequences and structures. Aim3: Validate deep learning-predicted TCR-antigen binding with retrospective public data and also prospective experiments. Significance and impact: Our deep learning models will revolutionize the design and implementation of multiple forms of immunotherapies, including, but not limited to TCR engineering. Our artificial intelligence-guided approach will improve the efficacy, shorten the time and lower the cost of the generation of these immunotherapies by several orders of magnitude, for multiple cancer types. This has the potential to dramatically benefit cancer patients by providing affordable individualized immunotherapies.

Read Less