A Network Based Approach for the Identification of Stress Responses In Lead Compound Transcriptome Profiles
Daniel R. Hyduke1, Heng-Hong Li1, Hyeon Ung Park2, Sean P. Collins2, Simeng Suy3, Jiri Aubrecht4, and Albert J. Fornace Jr.1. (1) Biochemistry and Molecular & Cellular Biology, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, 3970 Reservoir Rd. NW, Research Building E504, Washington, DC 20057, (2) Radiation Medicine, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, 3970 Reservoir Rd. NW, Research Building E504, Washington, DC 20057, (3) Drug Discovery Program, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, 3970 Reservoir Rd. NW, Research Building E504, Washington, DC 20057, (4) Global Research and Development, Pfizer, Groton, CT 06340
In addition to acting on a specific target, lead drug compounds possess the potential to initiate unexpected stress responses. Identification of these responses late in product development results in a significant loss on investment, and undetected effects may adversely affect the patient population. Additionally, the modes of action for promising drug candidates are often unknown, despite positive phenotypic effects. To facilitate the discovery of stress responses associated with lead compounds, we have developed a toxicogenomics database of the early transcriptomic response of a human lymphoblastoid cell line to a broad range of stress agents. Transcriptome measurements were taken for stress agent doses that elicited a robust response. The agents in this database include: a variety of anticancer compounds, endoplasmic reticulum stressors, antimetabolites, oxidizers, and other stressors. Because stress agents can induce a variety of overlapping responses, a straightforward comparison of a lead compound's transcriptome profile to the database will provide limited, if any, information about the compound's modes of action. To overcome this problem, we employed a biclustering approach to identify stable gene sets that are associated with the overlapping stress responses. These stress gene sets were assembled into a stress response network, and then the active stress components in a lead compound's transcriptome response were identified with network component analysis. We have validated our approach using independently generated transcriptome responses for positive and negative test cases. Additionally, we have used our approach to infer that a novel anticancer compound induces apoptosis by increasing oxidative stress--this inference was validated with independent biochemical assays.