Abstract 4265: Humanized 3D tumor models that are mutationally aligned with AACR GENIE patients predict IMM-1-104 activity in RAS-addicted tumors

Abstract Introduction: IMM-1-104, with pan-RAS activity through deep cyclic inhibition MEK, was evaluated in humanized 3D preclinical tumor models displaying diverse MAPK pathway activation events. Based on drug-response, sensitivity and resistance profiles, a biomarker signature for IMM-1-104 was developed in order to project potential therapeutic response of cancer patients found in the AACR Project GENIE (GENIE) database. Experimental Procedures: Humanized 3D preclinical models better predict in vivo tumor responses versus 2D culture and more accurately replicate biology of human tumors. Therefore, the antitumor activity of IMM-1-104 was evaluated in over 130 tumor models spanning 12 distinct histologies in the humanized 3D tumor growth assay (3D-TGA). Cell-based whole exome sequencing readouts were combined with 3D-TGA results to build a pharmacogenomic response algorithm. When applied to the GENIE patient database, resultant tumor-specific response landscapes helped to inform an early pan-RAS clinical trial design for IMM-1-104. Summary of New Data: A machine learning model was developed to predict IMM-1-104 sensitivity using response-associated genes and signaling networks that were identified using 3D-TGA pharmacogenomics data. This model was used to estimate GENIE patient IMM-1-104 response profiles across key solid tumor indications. In addition, mutation constellations from GENIE were compared with those observed in cell lines to identify preclinical models that best resemble real-world patients. This effort was designed to further enrich the translational fidelity of specific tumor models with the goal of translationally identifying patient populations most likely to benefit from IMM-1-104 treatment. Conclusions: The depth of response to IMM-1-104 was evaluated across a panel of diverse 3D-TGA tumor models and led to identification of a biomarker signature for therapeutically addressable MAPK pathway addiction. To translate these findings into a relevant clinical application, a response algorithm was developed and applied to the GENIE database, which has cataloged the molecular profiles of over 100,000 cancer patients. Mutational landscapes of patients within GENIE helped identify preclinical models that better represent patient profiles likely to be encountered in the clinic. This approach could, as a general principle, be applied as a tool for improving biomarker discovery and clinical translation of oncology drugs. Citation Format: Praveen Nair, Sarah Kolitz, Jason Funt, Peter J. King, Kevin D. Fowler, Anna Travesa, Ian Rose, John Brothers, Amy Axel, Scott Barrett, Benjamin J. Zeskind, Brett M. Hall. Humanized 3D tumor models that are mutationally aligned with AACR GENIE patients predict IMM-1-104 activity in RAS-addicted tumors. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 4265.