Abstract 109: Network-based model to identify potential therapeutic targets in breast, colon, and lung cancers

Abstract

Abstract Purpose: It is difficult to prioritize potential therapeutic targets from thousands of differentially expressed genes identified by genome-wide gene expression profiling in cancer. The vast array of in silico resources currently available in life sciences research offer the possibility of aiding drug discovery process. Here we propose to take advantage of these resources to develop a genetic network-based model to comprehensively and effectively identify potential therapeutic targets in several cancer types. Method: A whole-genome genetic network, which can reveal the tendency for genes to operate in the same or similar pathways, is first constructed from heterogeneous data using a developed machine learning approach, RVM-based ensemble. A tumor-specific network can then be generated by mapping the differentially expressed genes in a tumor to the whole-genome network. Finally, potential therapeutic targets can be identified as hub genes that are functionally associated to multiple existing cancer pathways in the tumor-specific network. Result: Here, the approach is applied to Breast, Colon, and Lung Cancer separately. In each case, differentially expressed genes are all ranked based on the extent of their functional association with multiple known cancer pathways in the tumor-specific network. The result in each case shows that higher ranked genes are cited by more literature respectively related to the three cancers (Spearman's Rank Correlations, R>0.2 with p<1×10−10); that is, they likely play more important roles in these cancers, compared to lower ranked genes. While mapping the results to gene annotation, we find that many kinase, receptor, and transcription factor related genes, which are often proposed as possible molecular targets, are ranked highly in all cases. We also find that the effective targets detected by siRNA screens tend to be ranked highly in each case (the area under the ROC curve, AUC>0.75). Additionally, we also identified drugs and compounds that can target the highly ranked genes based on known drug-target information. Targets of many drugs, already in clinical trials and used for treatment of the three cancers, are all highly ranked in each case. Other drugs and compounds identified but not in clinical trials have also shown anti-cancer effect and could be considered as potential novel drug for these cancers. Moreover, we also find several novel targets in each case, which are not yet identified as cancer genes, are highly ranked and also increase cancer cell death in siRNA screens. One example is CSNK1G2 (casein kinase 1, gamma 2) in Colon cancer. Conclusion: Our approach has demonstrated the ability to identify potential therapeutic targets in cancer systematically and comprehensively using integrated functional genomic and proteomic data. It also implies that the proposed approach could be utilized to generate personal therapeutics. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 109.