Construction of Gene Regulatory Networks based on Gene Ontology and Multi-variable Regression

Gene regulatory network in one of most important methods for understanding biological functions of genes from gene expression profiles. The effective clustering algorithm is the foundation for constructing gene regulatory network. In this paper, by introducing the concept of gene ontology (GO), which express the biological knowledge and has already been obtainable from biologists, a new distance metric is proposed to replace the conventional one which considers only the difference between genes in expression levels. Firstly, the corresponding parents of a target gene is determined by using the clustering technique based on the proposed distance between genes. Secondly, a multi-variable nonlinear regression is utilized to model the predictors for each target gene. Finally, for each target gene, coefficient of determination (CoD) is employed to compute the probability of selecting different predictors for each gene. Some experiments were performed for testing the constructed gene regulatory networks. The results show the effectiveness of the proposed method