Abstract 363: Planar filtered gene regulatory networks in breast cancer

Abstract

Abstract Gene co-expression network analysis has been shown to be effective in identifying functional co-expressed gene modules associated with complex human diseases such as cancer, Alzheimers’ disease, obesity and diabetes. However, existing techniques to construct co-expression networks require some critical prior information such as predefined number of modules to be detected and numerical thresholds for defining coexpression/interaction, or do not naturally reproduce the hallmarks of complex systems such as scale-free degree distribution of small-worldness). In order to mitigate these problems as whole or in part, we have developed a novel co-expression network analysis approach, Planar Filtered Network Analysis (PFNA). PFNA utilizes a graph filtering technique to identify “true” interactions by means of embedding candidate interactions on a topological sphere(2, 3). Planar Filtered Networks (PFN) are naturally scale-free, small-world, and are comprised of a number of highly co-expressed gene modules). Furthermore, PFNA allows differential topology analysis of networks from different disease states based on centrality and peripherality metrics). We performed PFNA on the breast cancer data from The Cancer Genome Atlas (TCGA), and identified a number of novel gene modules associated with overall survival of the whole cohort or patient subgroups defined by receptor status (ER, PR, HER2) and PAM50 biomarkers(6). PFNA uncovers not only gene modules enriched for genes in well-known cancer pathways such as cell-cycle and immune response, but also novel modules that prognostically stratify patients with triple negative or basal breast cancers. Notably, several modules involved in G-protein coupled receptor signaling, protocadherin α and β pathways are highly predictive of survival for patients with triple negative disease. Furthermore, the differential topology analysis reveals that breast cancer biomarkers such as PAM50 and poor prognosis gene sets developed by van’t Veer et al(7) are more central in the tumor network than the one based on the matched adjacent normal breast tissues. In summary, PFNA reveals a number of novel high-level molecular features of the highly heterogeneous breast cancer. This novel network analysis method provides a set of unsupervised tools to objectively identify subnetworks that are associated with complex diseases such as breast cancer. 1. R. Albert, A. L. Barabasi. Rev Mod Phys 74, 47. 2. M.Tumminello, T. Aste, T. Di Matteo, R. N. Mantegna. Proc Natl Acad Sci U S A 102,10421. 3. T.Aste, T. Di Matteo, S. T. Hyde. Physica A 346, 20. 4. W.M. Song, T. Di Matteo, T. Aste. Phys RevE Stat Nonlin Soft Matter Phys 85,046115. 5. F.Pozzi, T. Di Matteo, T. Aste. Adv Complex Syst 11, 927. 6. J. S. Parker et al.. Journal of clinical oncology: 27, 1160. 7. L.J. van 't Veer et al.. Nature 415, 530. Citation Format: Won-min Song, Tao Huang, Seungyeul Yoo, EunJee Lee, Yongzhong Zhao, Li Wang, Zhidong Tu, Xudong Dai, Hanna Irie, Jun Zhu, Bin Zhang. Planar filtered gene regulatory networks in breast cancer. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 363. doi:10.1158/1538-7445.AM2014-363