The Research Of Gene Regulatory Networks Based On Genetic Algorithm And Mutiple Time Delay Of Bayesian Networks

With the development of microarray technique, a lot of gene expression data come into bring, which makes it possible to know and excavate gene regulatory relationship. The purpose of gene regulatory networks is to discovery gene regulatory relationship from gene expression database seemingly disorderly and unsystematic, which provide reliable gist for scientist to understand biology cell cycle and protein function from the whole angle. This paper adopts gene expression sequence and use dynamic Bayesian networks to reconstruct the gene regulatory networks. And then this paper apply genetic algorithm to optimize the network structure, which improve the precision.First, after researching the base dynamic Bayesian networks, this paper mostly does two things: Combines the binding sites as prior knowledge, which improve the prediction prevision; it is permitted that different regulatory genes regulate the same target gene with different time delay in the same time of day when computing the fitness of ML and MDL. That make the model more agree with the real life cycle.Afterward, this paper applies the genetic algorithm to optimize the gene regulatory networks. A few of improvements is done: A total gene regulatory network is regarded as an individual of the genetic algorithm; the time delay computed according to the mutual knowledge is used to initialize the individual of genetic algorithm, which speed the convergence of the genetic algorithm. Decrease the rate of crossover and mutation to boost up the stability of anaphase. Apply binding sites to carry through binding sites guided mutation to optimize the precision. Select in random a pair of genes to mutate randomly to avoid over fitting.At last, Matthews correlation coefficient (MCC), normalized mutual information coefficient (MIC) and specificity (Sp) are used to evaluate the algorithm. Experiment prove applying genetic algorithm to optimize can get favorable result.Based on the upper idea, this paper achieves a platform of gene regulatory network's reconstruction to make the regulatory relationship and algorithm evaluation visual and the platform has practicability.