Research On The Reconstruction Of Gene Regulatory Network Based On Differential Equations Network Model

With recent advances in cDNA and oligonucleotide microarray technologies, it hasbecome possible to measure mRNA expression levels on a genome-wide scale. And manygene expression data can be got from biology labs. Bioinformatics methods can be usedto select the gene expression data, find the regulatory genes, and predict thecorresponding network, which is an important research content. Two parts were used inthe algorithm, one is the clustering, which is used to get some number of genes, anotheris the reverse engineering methods, which is used to get the regulatory networks.The algorithm still has some shortcomings that need to be improved, such asoperation time, insufficient, and so on. In this paper, the above problems are studied. Themain research content and innovation are as follows:(1) Based on multi source data fusion of biclustering algorithm PGABGenerally, gene expression data are described as a two-dimensional matrix form.one represents the gene, and the other represents conditions. The matrix with arbitrarydata represents gene expression level. This expression levels are usually derived from aspecified conditions of mRNA relative abundance. The goal of this paper is to propose ageneral biclustering methods, this method can be spotted microarray data of some localexpression data. The algorithm is applied to the Stanford University yeast cell cycleexpression database Spellman made the experiment of yeast gene expression data fordouble cluster, and with the traditional genetic algorithm optimization of the doublecluster results are compared, have verified the algorithm in double clustering accuracy onthe advantage.(2) SVD framework based on the differential equation model for constructing generegulatory networkBiclustering algorithm can generate the small scale target gene groups results forconstructing gene regulatory network. Differential equation model is used in this paper,but this model need search the solution space. The range is too large, all of them satisfythe equations of a real space conditions, thus singular value decomposition method wasused to reduce solution space. Singular value decomposition method was used and thelinear model can be derived in reverse engineering gene regulatory networks. Althoughthe singular value decomposition method can provide a very useful data intensivedescription method, but the operator alone operation may not be very accuratelypredicted the connection matrix. So it does not predict gene network behavior. The first use of the singular value decomposition method to construct the candidate solution set,and then through the corresponding data regression analysis to identify this solution ofthe sparse matrix. In every network weight matrix are SVD general solution validation, asby then accepted, otherwise calculated again. Through the simulation data and real dataresults, and compared with other traditional differential equation model algorithm, toverify the algorithm in shortening the operation time and improve the accuracy of theresults of execution efficiency.(3) OPSO method based on the differential equation model for constructing generegulatory networkAn improved particle swarm optimization is proposed in this paper, which can beused to infer gene regulatory network. One advantage is that the modified algorithm cansearch the best value of given function. Furthermore, the optimal algorithm can be usedto find the best suitable gene regulatory network of given genes. This method can alsoavoid local convergence in the optimal algorithm mostly. The simulated data and real datawhich is Yeast Saccharomyces cerevisiae cell cycle gene expression profiles from theSpellman’s data were used to test the efficiency of the proposed algorithm. And wecompared our algorithm with traditional genetic algorithm and particle swarm optimationalgorithm. The efficiency of the algorithm is demonstrated in this paper.