Cuckoo Search And Its Application On The Biclustering Analysis

Gene expression data reflect the expression level of thousands of genes under multiple experimental conditions which is obtained by direct or indirect measurement, such information about the change of genes expression, the correlation between genes and the impact of condition on genes can be found through the analysis of these data. They have important applications in many fields such as the diagnosis of clinical symptom, the judgments of drug efficacy and the reveal of disease mechanisms. As the wide variety of cells in organism and the spatial and temporal specificity of gene expression, the gene expression data is more complicated and grow faster compared with the genomic data. Therefore, the analysis of the gene expression data has been a key and difficult branch of bioinformatics. As the clustering methods are generally based on the entire attribute of data and can only locate the global information instead of the local information in the dataset, but a huge number of biological information is hidden in the local information. Therefore, the concept of biclustering is proposed in order to search the local information in the dataset more effectively.The cuckoo search algorithm is applied to the function optimization problem and the biclustering of gene expression data and we take the function to be optimized and the mean square residue of biclusters as the objective function respectively. The main contributions of this thesis include:1. The self-adaptive cuckoo search algorithm is proposed to solve function optimization problems on the basis of the cuckoo search algorithm. The self-adaptive version algorithm takes benchmark functions as its objective functions and five improvements are made to the cuckoo search algorithm which are:(1)The initialization method used in scatter search is adopted to initialize the population of solutions in order to make the solutions of the initial population can be more uniformly distributed in the search space.(2) The solutions with worse fitness values are accepted with a certain probability in order to increase to diversity of the population.(3) The randomly generation method is adopted to assign values to these solutions whose values exceed the border.(4) The mutation operation method is applied to the optimal solution of the current population in order to increase the ability of exploitation.(5) More search strategies and the corresponding dynamic adjustment mechanism are added to in order to make the self-adaptive version can adopt different search strategy according to different search strategy.2. The modified cuckoo search algorithm is proposed to solve the biclustering of gene expression data on the basis of the cuckoo search algorithm. The modified version algorithm takes the mean square residue of biclusters as its objective function and two improvements are made to the cuckoo search algorithm which are:(1)The matrix of gene expression data are divided into several parts with equal volume in order to generate biclusters in these parts with equal number.(2) The search step are associated with the number of row and column of the bicluster which the current search operation are based on in order to adjust the search step for different biclusters.