Study Of Intelligent Modeling Method Of Scaling Gene Regulation Network For Slime-forming Bacteria

Slime-forming bacteria is one of the main scaling microorganisms in the heat exchanger system.With the ability to rapidly multiply,they have a short growth cycle and can accumulate in the heat exchanger equipment quickly to produce a large amount of metabolites,thereby forming microbial fouling.The existence of microbial fouling not only greatly reduces the efficiency and the economic benefits of heat exchanger,but also increases the safety hazard of the heat exchanger equipment,which has great harm to the heat exchanger system.So far,as for the scaling mechanism of slime-forming bacteria,there has not been a unified mathematical model that can be used for experimental research,but the process of scaling is inevitably regulated by the scaling genes.Therefore,to construct the gene regulation networks of the scaling genes for slime-forming bacteria and study the regulation relationship among the scaling genes are of great significance for exploring the scaling mechanism of slime-forming bacteria and developing new methods to control microbial fouling.In this thesis,the key scaling genes of slime-forming bacteria were taken as the main study object,and data-driven method was served as the main study method.We have developed a new method to construct a gene regulation network,and completed the construction of the scaling gene regulation network for slime-forming bacteria.Firstly,in order to find the scaling genes of slime-forming bacteria,we have designed fouling inhibition experiments using high-frequency electromagnetic field and transcriptome gene sequencing experiments.By comparing the differential gene expression of slim-forming bacteria under normal condition and high-frequency electromagnetic field environment,forty-three scaling genes were found.We have made further efforts to find twenty key scaling genes and obtain gene expression data for four moments.Then,based on the comprehensive analysis of various gene regulation network models and modeling methods,we have proposed a new gene regulation network construction method based on adaptive long short-term memory neural network(LSTM),which consists of three steps:(1)LSTM is used to construct the gene expression level prediction model,and particle swarm optimization(PSO)is used to optimize the weight and the number of hidden layer nodes of LSTM;(2)Newton interpolation method is applied to expand the expression data of related genes;(3)Pearson correlation coefficient analysis method is applied to extract the regulation relationship among the expanded expression data of the related genes and the predicted expression data of target gene.In order to test the accuracy and universality of the proposed modeling method,we carriedout modeling analysis on the yeast cell cycle gene dataset and E.coli gene repair network dataset,respectively.The results show that the proposed gene regulation network construction method based on adaptive LSTM has high modeling accuracy.On both real biological datasets,the modeling accuracy and F-score are over 0.9,and the performance metrics are all superior to other modeling methods.Finally,we used the proposed modeling method to construct three gene regulation networks of twenty key scaling genes for slime-forming bacteria.