Abstract 822: OncoLoop: Closing the loop between patient-centered drug discovery and preclinical testing in precision-oncology

Abstract

Abstract Millions of individuals in United States are affected by Prostate Cancer (PCa), with tens of thousands dying every year, as therapeutic options for more advanced stages are not yet curative. A lack of suitable models of human cancer and inter-patient tumor heterogeneity often hinder the drug discovery process. In fact, the lost-in-translation phenomenon is one of the major causes of failure for compounds in the transitioning process from the workbench to the clinic. These limitations implicate considerable costs for the society in terms of both financial assets and human lives. To accelerate the drug discovery process, we devised and experimentally validated OncoLoop, a data-driven framework for the identification of clinically-relevant compounds that can be validated in preclinical models that optimally match patient-specific tumor dependencies. Specifically, we generated a set of Genetically-Engineered Mouse Models (GEMMs) of prostate cancer that recapitulate a wide variety of established prostate-cancer driver mutations. Their RNA profiles were then used to reverse-engineer a murine prostate cancer regulatory network, which was used to assess the activity profile of thousands of GEMM-specific regulatory proteins using the VIPER algorithm. We then assessed the match between the GEMM repertoire and several human prostate cancer databases–including from The Cancer Genome Atlas (TCGA) and Stand-Up To Cancer (SU2C) consortia–based on GEMM vs. patient Master Regulator (MR) protein conservation. This allowed selecting an optimally matched GEMM model, for virtually every patient in these databases, to serve as an in vivo surrogate model for screening patient-relevant drugs (i.e., a patient avatar). We then leveraged large-scale perturbational databases, comprising RNASeq profiles of 2 prostate cancer cell lines treated with >400 drugs, to prioritize drugs presenting the greatest ability to reverse MR protein activity in both the patient and in its optimally-matched GEMM, using the OncoTreat algorithm (Alvarez, et al. Nat Genet 2018). This formed a closed loop between each patient, its optimally matched GEMM and the drug(s) inferred to invert the activity of MR proteins in both the patient and the GEMM. OncoLoop selection of candidate drugs was validated in vivo. Notably, four out of five drugs were validated in the GEMM, including two inducing complete abrogation of tumor growth and two inducing significant tumor size reduction. In summary, we propose that given a patient's tumor biopsy, OncoLoop can identify an optimal GEMM model for in vivo validation and dramatically reduce the combinatorial search space of candidate drugs to a handful that may be effectively tested in that model. OncoLoop can be applied to any other cancer for which clinically-relevant tumor models are available. Citation Format: Alessandro Vasciaveo, Min Zou, Juan Arriaga, Andrea Califano, Cory Abate-Shen. OncoLoop: Closing the loop between patient-centered drug discovery and preclinical testing in precision-oncology [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 822.