Gaussian Graphical Model Based Metabonomics Network Analysis And Its Application

Metabolomics mainly studies the changes of endogenous metabolites in the body under the action of internal and external stimuli.The traditional metabolomics data analysis method only pays attention to the change of the concentration of a single metabolite,ignoring the correlation between metabolites.With the development of complex network theory and the deepening of system biology research on network analysis methods,it provides a certain theoretical basis for us to carry out metabolomics research based on network analysis.The biological metabolic system can be characterized as a network with metabolites as nodes and the correlation between metabolites as edges.In disease-related research,the metabolite correlation network model of the disease group and the healthy control group can be constructed separately,and the difference network analysis,The discovery of disease-specific sub-networks provides new ideas for metabolomics research.This article takes colorectal cancer as the research object to study the metabolomics network analysis method.It mainly involves the construction of metabolite association network in metabolomics and the analysis of difference network.The specific description is as follows:(1)We use the Gaussian graph model(GGM)method based on the l1-norm constraint to construct the metabolite correlation network model.This method can effectively measure the independent correlation between metabolites,even when the sample size is less than the number of variables(n <d)Under the conditions,it can also be solved;in addition,the model parameters are optimized by combining the St ARS method to obtain a relatively stable sparse metabolite correlation network.The above method was used in the modeling of human serum metabolomics data network,and the metabolite association network models of healthy people,polyps and colorectal cancer groups were constructed respectively,and the three networks were compared and found,from healthy to polyps and then In the colorectal cancer group,the network showed a gradual sparse trend,that is,as the disease progresses,the correlation between metabolites gradually decreases.(2)We compared and analyzed the existing differential network analysis methods,and proposed a direct difference differential network(d DGGM)modeling method based on D-trace for metabolomics data analysis,and used the ADMM method to derive the objective function solution process.This method is used to analyze the metabolomics data of colorectal cancer,construct a differential metabolite association network for the healthy group,polyp group and colorectal cancer group respectively,and count the number of nodes and edges of the three differential networks,find polyps group and healthy group and the differences between network edge of node number is far lower than the other two differences between the network that is,the metabolic difference between the polyp group and the healthy group is smaller than that of the colorectal cancer group and the healthy group,and the colorectal cancer group and the healthy group.Polyp group.