Abstract 3509: Using generative AI for filtering and comprehension of drug target discovery screen results

Abstract

Abstract Recent advancements in generative AI, including models such as GPT-4 and LLAMA2, have been rapidly integrated into cancer research. Cellformatica, Inc. has developed a pipeline designed to enhance the accuracy of functional and genomic drug target discovery screens by reducing the incidence of false positives and refining the interpretation process. This pipeline is capable of conducting comprehensive investigations, including genome-wide analyses, and is accessible via web and API interfaces. At the core of the pipeline is the use of the extensive global corpus of scientific literature to assess the relevance of each gene within the context defined by the experimental design of the screen. To address the issue of AI-generated inaccuracies, often referred to as 'hallucinations,' the pipeline incorporates a custom-tuned language model (LLM) component within the generative AI framework. This fine-tuning process is specifically tailored to the gene set under investigation. The pipeline's multi-agent architecture is a distinctive feature. It scores candidates based on three orthogonally-tuned criteria: effectiveness, confidence, and novelty. In benchmarking tests, this multi-agent scoring system has outperformed GPT-4 and has proven particularly adept at identifying candidate targets that may have been overlooked in previous research. These candidates are recognized for their significant therapeutic potential and are corroborated by existing scientific findings. The versatility of the Cellformatica pipeline is demonstrated through its application in three distinct areas of cancer research. First, it has been employed to optimize chimeric antigen receptor (CAR) T-cell therapies, a form of immunotherapy for cancer treatment. Second, the pipeline has been used to elucidate the network of chemotactic molecules that orchestrate immune cell infiltration and tissue remodeling in the lungs of patients with COVID-19. Third, it has been used to spatially define signaling pathways in humanized xenograft murine tumor experiments, for which existing cross-species database knowledge is limited. These applications underscore the pipeline's potential to contribute to the understanding and treatment of complex diseases. Citation Format: Andrew Rech, Yury Goltsev, Nikolay Samusik, Anna Kaznadzey. Using generative AI for filtering and comprehension of drug target discovery screen results [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3509.