Detection of Multi-Level Hierarchies in Multi-View Cancer Networks, with Applications to Glioblastoma Multiforme

Biological and other complex networks are generally believed to be hierarchically organized. High-throughput molecular sequencing technologies now make it possible to reconstruct large-scale biological networks based on different types of data at the genomic, transcriptomic, epigenomic, proteomic and metabolomic levels. A crucial task is to effectively integrate and analyze these different “views” of the data to gain a systems-level understanding of biological components, processes, and their functions. We here review recent advances in molecular data integration, multi-view learning and multi-level hierarchical community detection in big data networks. We then propose a novel method that integrates multiple views of similarities between data points into a single network via a diffusion process, and detects communities on multiple levels of hierarchy. On simulated data, we show that our approach is indeed able to capitalize on both common and complementary information contained in multiple views for the identification of an underlying multi-level hierarchical community structure. We apply our method to gene expression, copy number aberration and DNA methylation data from Glioblastoma Multiforme tumor samples to identify groups of genes that are highly co-regulated during disease progression. We verify that the resulting community structure is indeed representative of biological function by identifying various communities in which genes associated to known biological processes are highly overrepresented on statistically significant levels. We visualize the resulting network based on its multi-level hierarchical structure to allow for easy, intuitive exploration of the data.

Keywords: systems biology, cancer, data integration, multi-view data, network fusion, community detection, multi-level hierarchy, overrepresentation analysis, complex networks

Recommended citation: Arndt, P. (2018). Detection of Multi-Level Hierarchies in Multi-View Cancer Networks, with Applications to Glioblastoma Multiforme [Masters Thesis]. https://fliphilipp.github.io/files/MScThesisPhilippArndt.pdf