Research On Early Diagnosis Of Type 2 Diabetes Based On Markov Random Fields

Diabetes is a chronic disease that is difficult to cure,but in the early stage of Type 2diabetes,there is little difference in blood sugar fluctuation among the normal population due to the compensatory mechanism of islet cells,by the time a diagnosis is made with the current test,it is unlikely that the patient will be able to heal on his own with proper diet and metformin.In order to explore the early dynamic changes of Type 2 diabetes,we analyzed the tissue-specific expression data of mice,and obtained the evidence of differential expression between adjacent time periods to characterize the dynamic changes of gene differential expression.This data set contains abundant spatio-temporal information,but it is difficult to make full use of it in the past research.We use a Bayesian neighborhood selection method to estimate the gauss graph model(GGM).Instead of estimating the network independently for each group,when there are multiple sets of data available,the joint estimation of the network can make use of the information shared among the groups,and can improve the estimation of each individual network.We specify Markov random field(MRF)in potential state to combine complex data with spatiotemporal structure,and calculate model parameters with Monte Carlo expectation maximization(MCEM)algorithm,the pathological changes of different organs and tissues at different time were considered.The key feature of our method is to consider both the spatial similarity of gene expression level and the time dependence in order to extract biologically meaningful results from the data.