EAGER: Development of a Hybrid Knowledge- and Data-Driven Approach to Guide the Design of Immunotherapeutic Cells

University Of Pittsburgh, Pittsburgh PA

Investigators

Abstract

Over the past decade, immunotherapy has rapidly become the "new pillar" of cancer treatment, utilizing and strengthening the patient's immune system to attack tumors. Chimeric antigen receptors (CARs) engineered on T cells, a type of white blood cells, have revolutionized the treatment of blood cancers and have shown promise for treating solid tumors, as well as auto-immune diseases and chronic viral infections. The main goal when engineering CAR T cells is to generate T cell phenotypes capable of effective and durable tumor clearance with increased anti-tumor cytotoxicity, T cell persistence, and lower exhaustion. CAR intracellular domains play a key role in converting antigen recognition into these anti-tumor effector functions. Selecting among a plethora of candidate receptor domains and ordering them on a receptor, to optimize their effect on cellular function, presents both tremendous opportunities and considerable design challenges. A large-scale systematic computational exploration and recommendation of CAR signaling domains has the potential to transform the field of CAR-based immunotherapy by producing novel CAR T cell behaviors leading to safer, more effective therapies. At the same time, such studies offer excellent interdisciplinary training, bridging synthetic biology explorations and fundamental biology knowledge with innovative computational approaches. To accomplish these goals, this EArly Grant for Exploratory Research (EAGER) will explore a radically different CAR T cell design methodology, a hybrid artificial intelligence approach that integrates experimental data, through data-driven learning and inference methods, with knowledge sources, through knowledge-driven mechanistic network assembly and analysis. This project will systematically study the steps in the receptor design pipeline, and their full automation: retrieval of relevant information from literature and pathway databases, intracellular T cell network assembly, knowledge-based constraint generation and use in data-driven deep learning methods. With these explorations, this project will determine the most effective methods to address the uncertainty and training runtime of previous approaches, while providing reliable recommendations and explanations of CAR designs. Ultimately, this project would advance the knowledge and contribute novel research strategies in synthetic biology, systems biology, biosensing, and immunotherapy. The outcomes of this project, evaluation of novel algorithms and methods, network simulation and analysis data, and mechanistic explanations of recommended CARs, will be open source and publicly available for the wide scientific community to examine, utilize, and reproduce. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.