Discovering Gene Co-Expression Modules Using Fuzzified Adjusted Rand Index

Abstract

Understanding the interrelationship among genes in a cellular system is fundamental to the investigation of cellular activities, because the interrelated genes are either functionally related, controlled by the same transcriptional regulatory process or generally take part in a common biological process, and most importantly are known to be co-expressed genes. Most latent Mtb genes have been discovered but their functions, interrelationship and correlations that will help to develop protocol (s) to tame the menace of tuberculosis disease at latency have not been fully uncovered. We have developed a computational technique called Fuzzified Adjusted Rand Index (FARI) to effectively discover the co-expressed genes from identified latent Mtb genes and perform functional analysis of the gene sets using an annotation database. FARI, a modification of Adjusted Rand index used to compare clustering results, is designed to analyze, establish and quantify the expression trend of two genes with different sample points. Rank matrix of all the genes in consideration is produced after each gene has been analyzed with others, and the rank matrix serves as the basis of the co-expression discovery. A synthetic gene expression dataset, the biological benchmark dataset (E. coli), and different set of genes containing latent Mtb genes from an experiment result were fed into the computational tool, and different gene sets (modules) representing co-expressed genes were discovered. The discovered gene modules from latent Mtb genes are used to uncover the hub genes and their molecular functions. We have been able to identify different co-expression network from this analysis and assign biological functional meanings to some of the important Mtb genes that emerge from the experiment. Also, discovering gene co-expression module births gene co-expression network, which is a preliminary step towards gene regulatory network discovery.