The inference of predictor set in gene regulatory networks using gravitational search algorithm

Abstract

The inference of Gene Regulatory Network (GRN) using the gene expression data is a growing field in bioinformatics and biological systems. As a matter of fact, the inference of GRN is crucial in order to predict the biological processes. In addition, it would be beneficial to determine the behavior of the processes in order to avoid the occurrence of some unplanned processes (disease). Inferring truly GRN requires the accurate inference of the predictor set. The process of predictor set inference consists of realizing the dependency of target genes and their potential predictors. Generally, the main limitations of an accurate inference of predictor set are the large number of genes, the low number of samples and the presence of noise in the gene expression data. This paper presents an accurate framework using Gravitational Search Algorithm (GSA) to infer predictor subset of each target gene in a GRN. In this work, one heuristic algorithm is utilized for each target gene independently. In each population, a mass presents the predictor subset related to the target gene. To generate the initial population per each target gene, instead of choosing predictors randomly, they are chosen using the Pearson correlation coefficient. The Mean Conditional Entropy (MCE) is used to guide GSA (as fitness function). Experimental results on biological data and comparative analysis including a recently method based on Genetic Algorithm (GA) for the same purpose, reveal that the proposed framework achieves superior accuracy.