ClustEx2: Gene Module Identification using Density-Based Network Hierarchical Clustering

With the fast accumulation of large-scale omic data in cancer, it is easy to get a list of seed genes associated with any clinical phenotype change or anti-cancer drug perturbation. However, it remains a challenging task to functionally interpret these genes. Many studies indicate that genes work cooperatively as functional modules in complex cellular processes. More biological insights can be obtained by clustering these genes and their closely interacting neighbors in molecular networks as gene modules. Unlike the traditional network community detection, the gene modules involve in the seed genes that carry context specific information and their closely connected neighbors in the static molecular networks. We developed a new method ClustEx2, which can identify gene modules based on a set of user-defined seed genes in a given gene network. The method formulates the module identification in a unified framework according to a density-based hierarchical clustering method. ClustEx2 can incorporate both network topology and context specific information of seed genes along with their interactions, such as differential expressions and co-expressions. Its performance was systematically investigated for a known biological process, tumor necrosis factor induced inflammation, and it also helped obtain potential biological functions by analyzing the anticancer drug response associated modules using the TCGA data.