Research On Biclustering Of Gene Expression Data Based On Swarm Intelligence

The advent of high-throughput gene microarray technology has generated a large amount of gene expression data.These data play a vital role in tracking biological processes and discovering genetic rules,and analyzing pathology.Generally,researchers use clustering to mine relative sets of genes.And then,researchers organize and analyze them biologically.However,due to the unique data structure of the gene expression data and the biological significance behind it,traditional clustering methods that tend to find global patterns are not good at finding clusters with local patterns that meet the requirements.Therefore,biclustering analysis that more suitable for the characteristics of gene expression data was introduced.At present,the research on the application of swarm intelligence algorithms to biclustering analysis still has some problems.On the one hand,it is caused by the shortcomings of the swarm intelligence algorithm.For example,it may fall into a local optimal.On the other hand,it is caused by not organically combining the characteristics of swarm intelligence with biclustering analysis,such as selecting appropriate evaluation indicators for single or multi-objective optimization.Based on swarm intelligence optimization algorithms such as cuckoo search algorithm,firefly algorithm and bacterial foraging algorithm,this paper conducts analysis and research on the biclustering of gene expression data from the aspects of algorithm combination and multi-objective optimization.This paper aims to solve the problems of poor biclustering quality and biological significance of the current biclustering algorithms.The main work of the paper includes:(1)A hybrid biclustering algorithm CSFAB(Cuckoo Search and Firefly Algorithm hybrid Biclustering)based on cuckoo search algorithm and firefly algorithm is proposed.Considering that the cuckoo algorithm and the firefly algorithm can be regarded as a complementary relationship,the former has a strong global optimization ability,while the latter has a faster convergence speed,so this article attempts to mix the two.First,an effective hybrid strategy was determined through experiments,andthen the global search ability of the cuckoo search algorithm and the fast convergence ability of the firefly algorithm were effectively combined.The CSFAB algorithm can significantly improve the search speed and range,and at the same time can jump out of the local optimal solution and find biclusters containing different genes,thereby improving the diversity of biclusters.Compared with CSB,FAB,and PSOB algorithms,experiments show that the quality and biological significance of CSFAB algorithm is better.(2)MOBFOB(Multi-object Bacterial Foraging Algorithm Biclustering)based on a multi-object bacterial foraging algorithm is proposed.Because biclustering analysis can be considered as a multi-objective optimization problem,this paper improves the traditional single-object bacterial foraging algorithm based on the characteristics of biclustering analysis of gene expression data,mainly include: 1)Determination of better solution when there is no dominant solution;2)Sort according to the number of times they are dominated in the population;3)Introducing externally dominant solution sets to increase diversity.This algorithm optimize the bicluster's quality evaluation indicators such as mean square residual and volume simultaneously by using multi-object bacterial foraging algorithm,and find the dominant biclustering solution set.The quality evaluation index and bio-enrichment analysis of the biclustering prove that the MOBFOB algorithm can effectively and quickly find the biclusters with significant biological significance.