Gene expression data and cancer correlation analysis by Emerging Pattern Based Projected Clustering

Abstract

Cancer studies are one of the hot topics in medical and bioinformatics domains. Scientists are using microarray technologies and data mining techniques to study cancer at molecular levels. Two data mining techniques, namely, pattern mining and clustering, are heavily used in the field of bioinformatics to analyze gene expression data. In this thesis, the basic problem in organizing the information from the gene expression data in an easy understandable way for the domain experts in the further knowledge discovery process are investigated. We have introduced the Emerging Pattern Based Projected Clustering (EPPC) approach to organize the gene expression data into meaningful clusters. We apply the ideas of the emerging patterns and projected clustering together to form emerging pattern based projected clusters for the biologists. The resulting clusters can be used in the cancer detection problem and the experiment results show that its classification performance is comparable with ORCLUS, the state-of-the-art clustering approach. With its strength in readability, we believed that the resulting clusters are useful for the domain experts in conducting further experiments and studies.