Application Of EM And EKF Algorithm In Modeling And Analysis Of Gene Regulation Relationship

With the development of biology,the functional analysis of individual gene has been difficult to satisfy the need of research on the laws of the life phenomenon,so gene regulation relationship focusing on the interrelation of genes is becoming a research hotspot.Gene regulatory networks reflect the dynamic interrelationship between genes in system perspective.Its modeling and analysis can help us understand the underlying mechanism of cell activity and disease genes,also can provide basis for study the pathogenesis and drug design of complex diseases from the system and s.At present,the study of gene regulation network is mainly focused on modeling algorithms at home and abroad,while the feature that biological process is continuous but the experimental data are discrete has not been taken into account,and the network dynamics analysis combined with specific regulation network has not been carried out.In this paper,based on biological data,yeast protein synthesis regulation network and p53 gene regulation networks are modeled and identified using Expectation Maximization(EM)algorithm and the continuous-discrete Extended Kalman Filter(EKF)algorithm,and then the characteristics of network are researched.The main research contents are as follows:1)Based on microarray data of yeast protein synthesis,the parameters and states of the dynamic,stochastic and noise-involved gene regulatory networks models are identified jointly using EM algorithm.Then the model is verified,and the stability,accuracy and feasibility of the model are analyzed.Results show that the EM algorithm can establish yeast protein synthesis network accurately,and the established gene regulation network is stable and has good prediction ability.2)Based on microarray data of p53 related genes,the early and late regulatory models of the p53 related gene multiple loops network exposed to ionizing radiation are established using EM algorithm.Then the robustness and response ability of the early and the late stage models are calculated using LMI toolbox.Results show that the robustness of the p53 related genes multiple loops network is slightly decrease while the response ability is increase after ionizing radiation.3)The nonlinear dynamic stochastic noisy delay model of p53-Mdm2 gene regulatory network is established.The parameters and states of the established model are identified jointly using continuous and discrete EKF algorithm.Results show that the model can accurately simulate the damped oscillation process of p53-Mdm2 regulatory networks after ionizing radiation;Adding different dose of ionizing radiation to the former model as the input of the p53-Mdm2 regulatory model,the model can simulate accurately the response process of the p53-Mdm2 network in the different dose of ionizing radiation.Besides,the response process is consensus with biological experiment results.In this paper the yeast protein synthesis network model,the p53 related gene multiple loops network model,and the p53-Mdm2 networks are established.The established models can accurately predict the response process of the networks.The analysis of network characteristics is consistent with experiment results.The methods proposed in this paper can provide effective methods for biological experimental modeling and parameter identification of system dynamics,and supply the foundation for the research of system dynamics.