Metabolite Concentration Prediction In Regulatory Architectures Of Metabolic Reaction Networks Using DEA

As a non parametric method, Data Envelopment Analysis(DEA)is applied to many fields. Based on metabolic network analysis in biochemical system, this dissertation studies the models and methods to predict the metabolite concentration.This dissertation consists of two parts. First, a modified C~2GS~2 model is used to predict metabolite concentration with different regulatory structures. The number of inequality constraints in the modified model is decreased, but its efficiency is maintained. Three groups of data were checked using DEA models. In the 1st and 2nd groups, the enzymes change between±20% and±500%, and between±10% of their reference values, without modifications in the regulatory structure. In the 3rd group, the enzymes change between±20% and±500% of their reference values, and the number of biochemical reactions is decreased. The predicted values are valid for the 1st and 2nd groups, and their combination, respectively. But the model fails for the data consisting of all the three groups. It is concluded that the model is only valid for data from reaction networks with the same regulatory structure, and this characteristic can be used as a reference to distinguish metabolites with different regulatory structures. Second, by classifying the decision making units (DMUs), the prediction efficiency of metabolites with the same regulatory structure was maintained and the computation load was saved by about 50%. The computation result with Matlab indicates that the modified model and method are feasible, and of practical value for the analysis of metabolic activity.