Abstract Drug combinations are highly desirable in oncology as they can improve therapeutic response and overcome drug resistance. However, responses can be context-specific, meaning there is a real need for large-scale studies to fully explore the drug combination response landscape. We present a screen of 109 drug combinations in over 750 diverse cell lines covering 41 hematological and solid tumor types. Each combination was screened in a full 7 × 7 matrix format, representing over 2.3 million individual data points, the largest such study to date. We took a novel approach to assess combination benefit. Growth inhibition and highest single agent (HSA) and Bliss excess combination heatmaps were generated for all >68,000 combination-cell line pairs, and each was scored for high activity (combination Emax>0.5) and combination benefit/synergy (HSA>0.1). Out of 4578 combination-cancer type pairs 489 met the filtering criteria. We prioritized combinations based on their efficacy and cancer type selectivity to minimize risk of potential tissue toxicity. We identified several known combination-cancer type pairs (e.g. MCL1 inhibitor AZD5991 plus BCL2 inhibitor venetoclax in AML) as well as novel indications for combinations already known to be active in a different cancer type (e.g. AZD5991 plus AKT inhibitor capivasertib in endometrial cancer or AURKB inhibitor AZD2811 plus venetoclax in DLBCL). We selected 4 combinations based on clinical need and validated them further in vitro and in vivo (selumetinib plus venetoclax or AZD5991, AZD2811 plus venetoclax, and capivasertib plus AZD5991). To understand how molecular context affects drug response, we also used GDSC tools ANOVA to perform over 5.4 million statistical tests to identify statistically significant associations between drug response metrics and multi-omics features including mutations, CNAs, gene expression and methylation. Associations were identified with five response metrics, covering both single agent and combination responses. We looked for significant associations across 21 different subgroups of cell lines and 6 molecular ‘baskets’ (TP53, KRAS, MLL2/KMT2D, PTEN, PIK3CA, and BRAF). We identified 11,611 biomarkers which met the significance criteria (p⇐0.001, FDR⇐10%, and both positive and negative Glass deltas >=1). Moreover, we used biomarkers of single-agent response to identify any combination Emax or Bliss associated biomarkers that could not be explained by the action of either drug. This resulted in identification of 1,631 ‘emergent’ biomarkers. In summary, our screen and pipeline have been designed to optimize preclinical interpretation with a focus on actionability, particularly by our unique approach, and to provide a valuable resource for exploration by the wider research community. As a result of this screen, we identified and validated novel combination-tumor types in multiple cancer types. Citation Format: Azadeh Bashi, Elizabeth A. Coker, Krishna Bulusu, Marta Milo, Claire Crafter, James Lynch, David Jenkins, Howard Lightfoot, Brandon Willis, Courtney Andersen, Omid Tavana, Kevin Mongeon, Jacob Gordon, Paul Smith, Simon Barry, Ultan McDermott, Lisa Drew, Susan Critchlow, Mathew J. Garnett, Jerome Mettetal. Large scale pan cancer drug combination screening to identify effective and actionable combinations and biomarker hypothesis. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5321.