Computer-Aided Design of optimized Chimeric Antigen Receptors for hematological malignancies (CADCAR project)

Chimeric antigen receptor (CAR)-T-cell therapy has achieved remarkable clinical response rates in relapsed/refractory B-cell malignancies. Unfortunately, the frequency of relapse remains high as a result of decreased cellular fitness, poor anti-tumor activity or a lack of persistence of the CAR-T-cell product. While the CAR architecture is a foundational driver of CAR-T-cell responses, its design is poorly understood, in particular for the structural hinge domain. Current hypothesis-driven workflows are low in throughput, expensive and laborious, and complicate pattern recognition. This project aims to define CAR hinge domain design rules by studying the relationship between hinge properties and CAR-T-cell responses in the context of hematological malignancies. We employ high-throughput cellular assays to phenotypically and functionally evaluate a large library of novel hinge domain candidates. A machine learning algorithm will be trained to correlate hinge domain characteristics with the obtained cellular outputs. We anticipate to create an algorithm that is capable of predicting superior hinge domains from a naïve set of candidates against a multitude of target antigens. We envision that in the future this model can be further trained to include other CAR domains and domain combinations to assist in further personalization of CAR-T-cell therapy.

Keywords:CHIMERIC ANTIGEN RECEPTORS, MACHINE LEARNING

Disciplines:Computational biomodelling and machine learning, Hematology, Adaptive immunology, Cancer therapy