Computational Intelligence In Bioinformatics Applications

With the Human Genome Project (HGP) accomplished, the main battle field of life science has turned from "structure genome"to "function genome". HGP and other sequencing projects bring explosive accumulation of biological sequences, which leads that traditional experimental methods can not meet the demand at all. Biological data have following characters: large quantity, with noisy pattern, lacking uniform theory. Computational intelligence methods are right good at dealing with problems of this kind. This dissertation firstly introduces bioinformatics from aspects of history, status, significance and research scope, then defines computational method and describes briefly four main computational methods: artificial neural network, hidden markov model, support vector machine and genetic algorithm. Because BP neural network is used most broadly and latter parts will use it to study some bioinformatic problems, chapter two introduces the development trace of BP neural network and deduces this algorithm in detail. Following parts are applications of BP neural network in a few important fields of bioinformatics. Chapter three discusses the classical problem of using BP neural network to predict protein secondary structure. Novel progress and bottle-neck problem in this field are also included. Chapter four applies Bp neural network into domain structure class prediction. Based on the same dataset, proposed method gives higher prediction accuracy. Gene selection and cancer classification based on microarray gene-expression data is a new hotspot in bioinformatics. We present an improved gene selection (feature selection) method and construct a BP classifier. Adopting open acute leukemia data, computer simulation result tells good classification performance: when 46 informative genes are selected, classification accuracy reaches 100%; when reducing the number of informative genes to 6, only one sample is mis-classfied. The last chapter summarizes this dissertation. Computational intelligence methods have a great deal of applications in bioinformatics, and the topics referred in this paper are limited. Good research results of chapter four and five embody the beautiful combination of computational intelligence methods and bioinformatics.