MICA: microRNA integration for active module discovery

Official URL: http://dx.doi.org/10.1145/2808719.2808733

A successful method to address disease-specific module discovery is the integration of the gene expression data with the protein-protein interaction~(PPI) network. Although many algorithms have been developed for this purpose, they focus only on the network genes~(mostly on the well-connected ones); totally neglecting the genes whose interactions are partially or totally not known. In addition, they only make use of the gene expression data which does not give the complete picture about the actual protein expression levels. The cell uses different mechanisms, such as microRNAs, to post-transcriptionally regulate the proteins without affecting the corresponding genes' expressions. Due to this complexity, using a single data type is definitely not the correct way to find the correct module(s). Today, the unprecedented amount of publicly available disease-related heterogeneous data encourages the development of new methodologies to better understand complex diseases. In this work, we propose a novel workflow Mica, which, to the best of our knowledge, is the first study integrating miRNA, mRNA, and PPI information to identify disease-specific gene modules. The novelty of the Mica lies in many directions, such as the early modification of mRNA expression with microRNA to better highlight the indirect dependencies between the genes. We applied Mica on microRNA-Seq and mRNA-Seq data sets of $699$ invasive ductal carcinoma samples and $150$ invasive lobular carcinoma samples from the Cancer Genome Atlas Project~(TCGA). The Mica modules are shown to unravel new and interesting dependencies between the genes. Additionally, the modules accurately differentiate between the case and control samples while being highly enriched with disease-specific pathways and genes.