A disease module detection algorithm for lung adenocarcinoma tumor network with significance of connections and network controllability methodology

Abstract

The protein phosphorylation modifications are important to protein activities and functions. It has been widely recognized that dysfunctional phosphorylation modifications are related to cancer. Specifically, some single amino acid variations could disrupt existing phosphorylation kinase-substrate relationships and create novel kinase-substrate relationships. Besides, numerous network-based methods have been proposed to identify meaningful disease modules, which are locally dense subnetworks. In this work, we proposed a new network clustering method to uncover disease modules, which are correlated with the specific disease, based on significance of connections instead of local density. Specially, we build a weighted tumor network of lung adenocarcinoma with kinase-substrate relationships, tissue-specific gene regulatory network, pairwise gene expression data and mutation data. With appropriate parameters decided by a machine learning method, our method identified 9 disease modules. We found that these disease modules could effectively discriminate tumor samples from normal samples. Some significantly important genes in these modules have been identified as target genes of drugs recently. Our results provide insights into the disease mechanism underlying, and help identify more target genes of drugs in the era of precision medicine.