Clustering: A Novel Meta-Analysis Approach for Differentially Expressed Gene Detection

Abstract

Analysis of gene expression data obtained from microarray experiments is helpful for various biological purposes such as identifying Differentially Expressed genes, disease classification, predicting survival rate of patients etc. However, data from microarray experiments come with less sample size and thus have limited statistical power for any analysis. To overcome this problem, researchers are now relying on a more powerful technique called meta-analysis, an integrated analysis of existing data from different but related independent studies. Microarray data reveal that genes are normally expressed in related functional pattern, which suggests using clustering as an alternative technique to group genes into relatively homogenous clusters such as Differentially Expressed and Non-Differentially Expressed. In this paper, we explore k-Means Clustering technique to perform meta-analysis of gene expression data for finding Differentially Expressed genes. Comparative analysis of k-Means Clustering technique is performed, and the results are validated by various statistical meta-analysis techniques, which prove clustering as a robust alternative technique for meta-analysis of gene expression data.