Protein kinases are major regulators of cellular signalling and known drivers of cancer and other diseases. In the last two decades, kinase inhibitors have gained much attention as cancer therapeutic agents. In this study, we combined drug vulnerability data of 19 breast cancer cell lines with kinase activity data, and employed machine learning and information theory to de-convolve cell line specific kinase addictions. The study was mainly focused on triple-negative breast cancers (TNBC), which still lack targeted therapy. With this approach, we identified several novel TNBC cell line selective kinase addictions (e.g., FGFR2 for MFM-223) and, as added validation, detected previously established kinase sensitivities (e.g., BRAF in DU4475, ERBB2 in luminal cell lines SK-BR-3 and BT-474, along with generic addiction towards cell cycle and apoptosis regulators like PLK1 and PI3K). A major benefit of this approach is that it can identify readily druggable targets while avoiding false positives and undruggable candidates from loss-of-function analyses. Comparison of kinase inhibitor vulnerabilities with loss-of-function data showed significant discordance, highlighting the need for cautious evaluation of loss-of-function data for drug target identification. Our study underscores the heterogeneity of drug response in TNBC and provides a platform for rapidly identifying vulnerabilities and guiding targeted therapy regimens.

Citation Format: Prson Gautam, Hassan Al-Ali, Krister Wennerberg. A chemical screening and machine learning approach to de-convolve kinase addictions in TNBC [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2017 Oct 26-30; Philadelphia, PA. Philadelphia (PA): AACR; Mol Cancer Ther 2018;17(1 Suppl):Abstract nr A149.