PO-023 A multi-step framework to analyse high-throughput drug combination screens

Abstract

Introduction Tailored, genotype-driven therapies are the future for treatment of heterogeneous cancers like breast cancer. However, despite their high therapeutic index, targeted agents often have limited success in single agent treatment setups due to variable clinical response rates and resistance. Thus, a framework to develop drug combinations for treatment of cancer is needed. Material and methods Our group has an ongoing effort to systematically screen two-drug combinations in cancer type-specific and pan-cancer high-throughput drug combination screens. We have screened over 1000 combinations, including both broad-acting chemotherapeutics and targeted compounds. Cancer type-specific drug combination screens were conducted in breast and colorectal cancer cell lines, using a panel of 50 cell lines in each screen. Acquired data has been used to determine a strategy to extract the most promising drug combinations in each tissue. Results and discussions We propose a multi-step analytical framework for the identification and prioritisation of synergistic combinations. Initial filtering is based on several parameters, including synergy (z-score), XMID and Emax. Robustness of drug response effects is assessed through rescreening of selected drug combinations. Further prioritisation of promising combinations in a tissue-specific setting is achieved through stratification of the combination drug response in cell line subgroups, including segregation based on commonly mutated genes and molecular subtypes (e.g. PAM50 subgrouping in breast cancer). Moreover, systematic exploration of genotype-drug synergy associations will be used to identify biomarkers of response for patient stratification. Conclusion In conclusion, we have developed a multi-step framework to identify clinically relevant synergistic drug combinations complementing current cancer therapy options. This pipeline may facilitate future combination drug synergy predictions and validation.