Abstract 3556: A pan-cancer ex vivo drug screen database for next-generation pharmacogenomics and functional precision oncology

• Published in: Cancer research (Chicago, Ill.) Vol. 84; no. 6_Supplement; p. 3556
• Main Authors: Markus, Anneliese, Pichotta, Karl, Quinn, Jeffrey, White, Jessica, Tosh, Christopher, Liu, Jinrui, Coyne, Erin, Tansey, Wesley
• Format: Journal Article
• Language: English
• Published: 22.03.2024
• ISSN: 1538-7445; 1538-7445
• DOI: 10.1158/1538-7445.AM2024-3556

Abstract Large-scale drug screens across thousands of drugs and biological models enable pan-cancer analyses that can reveal novel insights into therapeutic response and resistance. Existing databases, such as the Cancer Dependency Map only screen established 2D cell lines, which are known to poorly translate to patients. Ex vivo models such as patient-derived cells, spheroids, and organoids, have consistently demonstrated superior translatability from drug screening responses as they better replicate the original tumor. However, the cost, complexity, and difficulty of developing and screening ex vivo models has thus far prevented the establishment of any large-scale database comparable to existing cell line databases. We present the pan-preclinical (PPC) project, a large pan-cancer database of ex vivo drug responses comprising 2.1M raw dose-response measurements with curve-level summary statistics and metadata. The PPC database is assembled from over 30 previously published ex vivo studies and contains 2,880 ex vivo models spanning 106 cancer subtypes tested against subsets of 3,023 drugs. Compounds are harmonized across studies and paired with detailed chemoinformatics derived from PubChem. Each ex vivo sample is matched to an underlying patient and paired, where available, with genomics, transcriptomics, and clinical information such as sex, treatment history, and overall survival. Two thirds of the underlying studies did not originally make the underlying drug screening data available at publication time, making the PPC database a novel resource beyond data integration. Results: To investigate the power of the PPC database, we developed a preliminary computational modeling and statistical analysis pipeline. A hierarchical Bayesian factor model was developed to estimate dose-response curves, remove batch effects across studies, and impute missing drugs. Low-dimensional projections of ex vivo drug responses show biologically meaningful clustering. Statistical analyses of both observed and imputed drug responses recapitulate standard of care drugs as significantly effective across several cancer types. Overall, we envision the PPC project laying the foundation for the next generation of large-scale pharmacogenomic analyses and aiding in the coming era of functional precision oncology. Citation Format: Anneliese Markus, Karl Pichotta, Jeffrey Quinn, Jessica White, Christopher Tosh, Jinrui Liu, Erin Coyne, Wesley Tansey. A pan-cancer ex vivo drug screen database for next-generation pharmacogenomics and functional precision oncology [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 3556.