Abstract A067: Evaluation of large-scale quantitative systems pharmacology model predictions of in vitro cancer cell line sensitivity to TAK-659, an investigational oral spleen tyrosine kinase (SYK) inhibitor

Predicting the sensitivity of cancer cells and patient tumors to a drug remains a fundamental challenge in oncology drug development and patient care. Single-gene biomarkers of drug sensitivity ignore the crosstalk of gene aberrations seen clinically. In the case of the investigational oral spleen tyrosine kinase (SYK) inhibitor, TAK-659, we sought to evaluate computational methods that could eventually predict patient sensitivity to TAK-659 by first testing these methods on in vitro cell line sensitivity. Cellworks has built and curated a 500,000-reaction mechanistic pathway model of cancer that can predict sensitivity to an agent. The drug agent is modeled using its known mechanism of action and the patient model is derived from functional interpretation of genetic alterations. This computational biological model (CBM) represents phenotypes proliferation, viability, apoptosis, metastasis with detailed coverage of growth factor signaling, cell cycle regulation, tumor metabolism, oxidative stress, epigenetics, protein homeostasis and DNA damage repair. To test the ability of this technology to predict sensitivity to TAK-659, Takeda provided the names of 18 cell lines for which TAK-659 sensitivity had been tested but withheld from Cellworks. Based on publicly available aberrations (mutations and copy number variation [CNV]) in these cell lines and the mechanism of action of TAK-659, the model correctly predicted the in vitro sensitivity of 14 of these cell lines (78%). After unblinding and subsequent adjustment of a small number of model parameters, specifically the dominance of contribution of gene aberrations in context of the cancer indication, the model correctly predicted TAK-659 sensitivity of 17 of the 18 cell lines and more importantly 65 of 70 completely different cell lines in a blinded manner (93% accuracy, area under receiver operating characteristic [AUROC] curve = 0.879). For CBM prediction accuracy, the correctness of functional interpretation of genetic alterations and the corresponding dominance of the downstream propagation are key determinants. For low-prevalence gene mutations, the functional effects are mostly unknown. Nevertheless, based on the in vitro results for TAK-659, CBM represents a promising approach for predicting drug sensitivity. In the future, this work will be extended to test the predictive accuracy of CBM to distinguish between TAK-659 responders and non-responders in non-Hodgkin lymphoma using next-generation sequencing data from patients. Citation Format: Taher Abbasi, Shireen Vali, Miguel Williams, Jessica Sappal, Cong Li, Hyunjin Shin, Rick Gregory, Yujun Wu, Shining Wang, Kate Stumpo, Dean Bottino. Evaluation of large-scale quantitative systems pharmacology model predictions of in vitro cancer cell line sensitivity to TAK-659, an investigational oral spleen tyrosine kinase (SYK) inhibitor [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2019 Oct 26-30; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2019;18(12 Suppl):Abstract nr A067. doi:10.1158/1535-7163.TARG-19-A067