Abstract 2462: HemeBase: a comprehensive model-to-patient genomic map for AML and DLBCL

Abstract

Acute myeloid leukemia (AML) and Diffuse Large B-cell Lymphoma (DLBCL) are challenging hematological malignancies due to complex mutational patterns within and across patients and have unmet medical needs for therapies tailored for different disease subtypes. Translating AML and DLBCL preclinical findings into meaningful clinical outcomes can be inefficient and costly, as evidenced by failure of 85% of oncology drugs entering clinical testing to gain FDA approval and costs of bringing drugs to market exceeding $2 billion dollars. There is a pressing need for disease focused comprehensive knowledgebases to provide insight into whether preclinical models represent patient disease, assist in selecting accurate models to define proof of mechanism and advance therapeutic strategies. We provided Hemebase as the first comprehensive functional-genomic map for a one-stop-shop solution to enable real-time assessment of the relationships between thousands of patients and preclinical models within DLBCL or AML. Hemebase is an interactive, web-based tool to browse, compare, and visualize models and patients from the largest functional-genomic knowledgebase to date, including over 600 AML patients, over 1000 DLBCL patients, over 60 AML/DLBCL cell lines and sources of PDX mouse models (Crown/Champion). The multimodal data incorporated into Hemebase include somatic mutations, copy number alterations, expression values, pathway activity scores, drug screening data, and clinical features (i.e. diagnosis, previous treatment, and clinical outcomes). Importantly, several unbiased methods of clustering cross preclinical models and patients are incorporated, including a non-negative matrix factorization (NMF) that enables interactive assessment of model-to-patient similarity on molecular basis. NMF integration allowed us to recreate the methods used in the Schipp lab’s recent Nature Medicine paper to determine where cell lines lie within the the relevant 5 clusters of published patient data and perform real time analysis of how the clusters change based on varying the model inputs. To enable predictive biomarker discovery, clinically relevant filters can be simultaneously applied across models to analyze drug sensitivity - this was used to reveal the specific patient subpopulations that carry mutations that conferred drug sensitivity. These methods enabled the identification of an AML patient population that closely resembles previously used PDX models in terms of genomics and drug response, allowing for broader understanding of pharmacological response between these two models and a path for future investigation. Taken together, Hemebase provides unique opportunities to reveal the full landscapes of AML and DLBCL disease biology and therapeutics from enabling a comprehensive model-to-patient genomic map for AML and DLBCL. Citation Format: Kathleen A. Burke, Tyler Faits, Adriana E. Tron, Jay Mettetal, Andrew Bloecher, Zhongwu Lai, Jonathan Dry, Bolan Linghu. HemeBase: a comprehensive model-to-patient genomic map for AML and DLBCL [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2462.