The Analysis And Restruction Of Genetic Regulatory Networks Based On Evolutionary Computation Method

Fifty years ago, Watson and Crick identified the physical structure of theDNA, thus starting a new age for biological research. Since then, SystemsBiology has become an extremely important field in biology, which aims atdeep insights into biological systems. The progress of system biology revealsthe complexity of the interaction of mass genes, and thus traditionaldescriptive method in biology and analysis by disassembing faces the cruelchallenging. Over the years with the completion of "Human Genome Project",the appearance and application of DNA microarray technology make itpossible to abbtain quatitative measurement of expression of thousand ofgenes that present in a biological samlpe simultaneously and provides basesfor the research of the complex large-scale genentic regulations between genesby mathematical and computational method .Some researchers have began toprotract the genetic regulatory network in the whole living cell.Gene networksis the collection of gene-gene regulatory relations at the expression level .Because the genetic regulatory network is a dynamics model which hasnonlinear traits such as robustness ,hierarchy and so on ,the restruction ofgenetic regulatory network is to restruct the genetic interactional model thoughthe massive gene expression data combined with some analysis andcomputational method to simulate system dynamic behaviors,which can take asight of the inter-dependent relationships between genes.Contrarily,the modelestablished can direct the futher biologic experiments.Based on the crossing ofsubjecets on molecule biology,nonlinear maths and informatics science ,theanalysis and restruction of gene regulatory networks have been an importantresearch field in post_genome era. This paper reviews for the mostly mathematic method and models todescribe genetic regulatory systems that have been employed in SystemsBiology and bioinformatics ,such as clustering, directed graphs , Booleannetworks model,linear combination model ,weight matrices model ,mutual-information networks model,correlation coefficient model,Bayesiannetworks ,differential equations and so on.Every methods has their ownadvantage but also some limitation,there are no the best model for genenticregulatory system.Boolean networks model is mathematically trackable,and itssimplicity allows examination of large systems.Howerer,it can not infernetworks with feedback loops.Generally, continuous networks are used toanalyze the genetic regulatory network and reconstruct it. Weight matricesmodel is also applied to research the genetic network earlier. It can solve theproblems that whether there are interactional actions between genes anddescribe the strength according to the weight, but this approach has somedeficiency. For example, it can not describe the regulatory relation amongdifferent genes accurately. Linear model can also be used to find the networkmodel from the microarray data easily, but it is not realistic that we supposethe genetic regulatory relationships are linear. Another disadvantage is that thelinear model can only describe a single attractor. The information entropymethod describes a probability relationship, but the regulatory relationships itdescribes are not correct enough. Correlation coefficient model is a widelyused one at present, but the calculation of correlation coefficient is needed tobe researched and improved. Nowadays, one of the most widely researcheddynamic continuous modeling is S-system . An S system has a group ofspecial power law ODEs. This structure has powerful description ability tocatch the dynamics of a biology system. Because there have been manyanalysis and calculation methods for S-system, it have many advantages onsystem and control design. So in this paper we use S-system to reconstructgenetic regulatory networks .In this paper, we mainly check the approach using genetic algorithm (GA)to search the parameters set of S system, and validate the effective of thismethod,that is, it can predict the network which tallies to the experimental datahighly.However because differential evolution algotrithm is a new intelligenceoptimization method which is supposed recently, and has proved to have theexcellen()t characters of robustness, efficiency and convergence. So we alsoconstructs the genetic regulatory networks by DE with S-system. Theexperimental result shows that these two approaches can get the comparativelyaccurate network model, but the later performs better than GA on the speedand veracity. So as a new intelligence optimization method, it will have animportant application value on reconstructing regulatory network. Because thecomplexity of S-system and the noise of the true data , we use the simulatesdata instead of the true onesmany problems on genetic regulatory network are still not solves yet.ButWith the development of bioinformatics,more and more researchers will haveaware of the importance of this field and turn to it .So with moreunderstanding of bioinformatics , Human can get more genetic informationsfrom the genetic regulatory networks.