Abstract
AbstractIn this paper, we propose a novel approach namely MSIGNET to identify subnetworks with significantly expressed genes by integrating context specific gene expression and protein-protein interaction (PPI) data. Specifically, we integrate differential expression of each gene and mutual information of gene pairs in a Bayesian framework and use Metropolis sampling to identify functional interactions. During the sampling process, a conditional probability is calculated given a randomly selected gene to control the network state transition. Our method provides global statistics of all genes and their interactions, and finally achieves a global optimal sub-network. We apply MSIGNET to simulated data and have demonstrated its superior performance over comparable network identification tools. Using a validated Parkinson data set we show that the network identified using MSIGNET is consistent to previously reported results but provides more biology meaningful interpretation of Parkinson’s disease. Finally, to study networks related to ovarian cancer recurrence, we investigate two patient data sets. Identified networks from independent data sets show functional consistence. And those common genes and interactions are well supported by current biological knowledge.
Publisher
Cold Spring Harbor Laboratory
Reference23 articles.
1. Human Protein Reference Database and Human Proteinpedia as discovery tools for systems biology;Methods in molecular biology,2009
2. Chen X. , Xuan J. , Shi X. , Shajahan-Haq A.N. , Hilakivi-Clarke L. and Clarke R. (2013), 2013 IEEE International Conference on Bioinformatics and Biomedicine, pp. 16–18.
3. ChIP-BIT: Bayesian inference of target genes using a novel joint probabilistic model of ChIP-seq profiles
4. Banerjee S. , Chen X. , Wu X. , Xie H. , Xuan J. and Feng W.-c. (2017) ChIP-GMM: A Gaussian Mixture Model for InferringBinding Regions in ChIP-seq Profiles.
5. A Directed Protein Interaction Network for Investigating Intracellular Signal Transduction
Cited by
3 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献