SCHOOL OF COMPUTING, NUS DOCTORAL SEMINAR BY MR DONG DIFENG Computational techniques for drug pathway identification and disease treatment optimization Meeting Room 4(COM1 01-23) 24 January 2008, 1.30pm Abstract: Gene expression analysis techniques have been used for disease subtype diagnosis, disease subtype discovery, and treatment response understanding in the last decade. The research crux now is how to interpret a list of identified genes in biological context; or in other words, how to select genes with respect to biological meaning. Biological pathways contain potential information to answer these questions. Since gene products function by interacting with each other, pathways provide a benchmark for cancer researchers to select, rank, and evaluate genes against high-throughput expression datasets. In our current research, we design computational systems to understand drug treatment response in biological pathway context. Our target is to evaluate drug effect for individual organisms and provide directions for cancer treatment optimization. In this proposal, we introduce the background of our research, summarizing the achievements in disease subtype diagnosis, new subtype discovery, and treatment response understanding, with an in-depth review of expression analysis with biological networks. Recently, we have designed a drug pathway identification system for a nasopharyngeal carcinoma (NPC) study. In this study, 3 NPC cell lines and 13 NPC patients were treated with a cyclin dependent kinase (CDK) inhibitor, CYC202. As a result of the treatment, both cell lines and patients responded to the treatment differentially. Our system generates hypotheses for the regulated genetic pathways in response to the drug treatment, and identify the differentiation of pathway status between individuals. Interestingly, the evaluated pathway status are consistent between the two experiment groups, and perfectly separating the responders and non-responders in the patient dataset. Furthermore, we confirm our discovery with extra medical assays and publications in literature. Both results suggest the identifications of our system provides plausible hypotheses for further research.