Intelligent topological differential gene networks

Abstract

Microarray gene expression profiles are frequently explored to understand the causal factors associated with some disease. To date, most of the research being conducted is restricted upon comparison of expression values across more than one condition or the discovery of genes having altered interaction levels with neighbours across conditions. Therefore, differential expression (DE), gene correlation and co-expression have been intensively studied using microarray gene expression profiles. However, in the recent past the focus has been shifted towards conglomeration of differential expression and differential connectivity properties to gain a better insight of the problem, such as investigating the topological overlap (TO) of the network formed by DE genes using the generalized topological overlap measure (GTOM). In this work, we explore through the unweighted–TO networks which requires selection of a smart threshold to transform the GTOM structure into a differential network. The essence of our work lies in the generation of a series of GTOM threshold pairs across different conditions from which the best threshold pair for a network (across different conditions) is selected by comparing the cumulative effect of TO and p-value obtained from the series of threshold pairs.