Automatic Inference of Gene Regulatory Network Using Dynamic Model Based on Law of Mass Action

Abstract

One of the central tasks in gene regulatory network is to understand the interaction mechanism between genes and to further reveal the biological function of genes by analysis of the network. Since the procedure of the gene regulation is essentially composed of a series of elementary chemical reactions, and the law of mass action is often used to explain and predict the behaviors of chemical reactions in dynamic equilibrium, the mass action model can describe the interaction mechanisms more accurately. In the past, people often constructed this model manually. But it is a time-consuming work and requires very professional knowledge of biology. With the rapid growth of biological data, it would be unable to meet peoples’ need. Therefore, we proposed an automatic generation method for mass action model. The algorithm uses the evolutionary computation method such as Population-based Incremental Learning (PBIL) and Trigonometric Difierential Evolution (TDE) to learn the expression data, then automatically generates the network structure and identify the parameters of network simultaneously. Experiments carried on an artiflcial synthetic network with various levels of noise, as well as on two well-known reallife genetic network show that our approach can successfully auto-flnish the task of model construction and parameter identiflcation. Compared with other works, this method also has a great improvement in performances. Moreover, the parameters in this model have clear biochemical meanings and are beneflt to further analysis.