Abstract 5553: A computational model for integrating genomic data with public datasets for molecular tumor board recommendations

Abstract Recent genomic profiling studies in pancreatic adenocarcinoma (PDA) have revealed actionable mutations affecting multiple signaling pathways, but in spite of these mutations, targeted inhibitors of these pathways have low success rates. A possible reason for these failures is that single-gene biomarkers (e.g. a KRAS mutation as an indicator of MEK inhibitor sensitivity) fail to account for crosstalk within and between dysregulated pathways. We have previously curated a knowledgebase of published studies as evidence to support molecular tumor board recommendations to cancer patients after multi-omic profiling. Here we present a computational framework for integrating this knowledgebase with drug response data from cancer cell lines to propose “actionable” biomarkers based on a panel of pathways instead of targeting a single gene mutation. We constructed a computational model encompassing a broad range of cancer-related pathways, including RAS/RAF/MEK/ERK, PI3K/AKT, cell cycle regulation, and DNA repair. The model consisted of a set of ordinary differential equations (ODE) with protein interactions following Hill-type kinetics and the rate of cell division and apoptosis modeled dependent on key signaling nodes, including the level of phosphorylated ERK and AKT. We integrated two sources of publicly available data: 1) published studies correlating phosphoprotein measurements and resistance pathways to targeted inhibitors in clinical development; and 2) mutation data correlated with drug-specific response metrics (e.g. IC50 values), such as CCLE and NCI-60. We systematically screened frequently observed overlapping disrupted signaling pathways (i.e., combinations of mutations) by simulating predicted IC50 values for targeted inhibitors. Based on these simulations, we then simulated the effect of pairs of drugs to explore which drug combinations may be best suited for inhibiting tumor growth when tumors harbor multiple mutations. We present two applications of this computational approach: a comparison of CDK4/6 inhibition in CDKN2A-mutated PDA vs. hormone receptor-positive breast cancer and a comparison of PARP inhibition in BRCA1/2-mutated PDA and ovarian cancer. The predictions generated by our simulations were consistent with clinical observations in that fewer combinations of mutations in PDA were sensitive to these inhibitors than in breast and ovarian cancer, suggesting ways to refine biomarkers for sensitivity to these drugs in PDA. The computational approach presented here takes into account multiple datasets from a knowledgebase to provide a prioritized list of treatments that match a patient’s molecular profile while also providing the rationale for the recommendation. This represents a step toward incorporation of systems biology in precision oncology. Citation Format: R Joseph Bender, Edik Blais, Apoorva Kulkarni, Michael J. Pishvaian, David Halverson, Jonathan R. Brody, Emanuel Petricoin, Subha Madhavan. A computational model for integrating genomic data with public datasets for molecular tumor board recommendations [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 5553. doi:10.1158/1538-7445.AM2017-5553

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.