Abstract 4912: Systematic, comparative network analysis on non-small cell lung cancer

BACKGROUND: Most cancers lack any effective early disease markers, prognostic and predictive signatures. We fail treating cancer due to multiple ways cancer initiates and develops treatment resistance. While drug modes of action are complex and poorly understood, using Comparative Toxicogenomics Database is an effective way to identify drug combinations. Integrating signatures with deregulated network information may lead to identifying novel treatment option for individual patients. APPROACH: A systematic graph analysis was used to extract network structure differences between normal and tumor patient samples in non-small cell lung cancer. Three gene expression datasets with 27 squamous cell carcinoma, 129 adenocarcinoma, 20 large cell carcinoma and 141 normal samples, and 18 prognostic non-small cell lung cancer gene signatures were used to construct normal and tumor co-expression graphs. RESULTS: We enumerated all 5-node graphlets in normal and tumor graphs, and separated them into 3 categories: unique for normal graph, unique for tumor graph, or present in both. We further focused on subgraphs with the same membership across all 3 datasets and unique to tumor graph. Using gene enrichment analysis with a hypergeometric test we identified 9 subgraphs significantly enriched in the term “regulation of lymphocyte activation” (p