An improved identification technique of gene regulatory network from gene expression time series data using multi-objective differential evolution

Gene regulatory network provides the knowledge of interaction strength among the genes in living organisms. Accurate identification of gene regulatory network is of prime interest to the researchers in recent time. Different researchers applied different optimization techniques to solve this problem. Most of these optimization techniques considers the square error between reference and simulated gene expression as their objective and minimize it to get a solution for the identification problem under consideration. But these techniques do not guarantee a unique set of network parameter, because the squared error is a non-linear multimodal surface of network parameters. Therefore considering only square error as the objective function is not a good choice. An alternative way of formulation of this problem is to validate it from different perspective. In this paper, we propose a technique for identification of gene regulatory network using multiple objectives. The objectives are designed to make the identification technique more robust. Multi-objective differential evolution is used to find a set of pareto-optimal solutions with respect to the objective functions. Among those solutions, one is chosen according to some suitable criterion. Computer simulation has shown that the proposed technique can identify useful interaction information from gene expression time series data.