Autotuned Multilevel Clustering of Gene Expression Data

DNA microarray technology has revolutionized biological and medical research by enabling biologists to measure expression levels of thousands of genes in a single experiment. Different computational techniques have been proposed to extract important biological information from the massive amount of gene expression data generated by DNA microarray technology. This paper presents a top down hierarchical clustering algorithm that produces a tree of genes called GERC tree (GERC stands for Gene Expression Recursive Clustering) along with the generated clusters. GERC tree is an ample resource of biological information about the genes in an expression dataset. Unlike dendrogram, a GERC tree is not a binary tree. Genes in a leaf node of GERC tree represent a cluster. The clustering method was used with real-life datasets and the proposed method has been found satisfactory in terms of homogeneity, p value and z-score.