Classifying Microarray Data Using Pairwise Similarity between Gene Profiles

Figure 1: Dendrogram with a threshold applied (---) to create clusters. The first cluster has genes A and B while the second one has genes C, D, and E. genes to be measured simultaneously. However, interpretation of the large data sets is complex. Commonly used methods include hierarchical clustering and k-nearest-neighbor [3]. After applying these methods, clusters of co-expressed genes are produced, which can then be used as a fingerprint to identify a biological response. The basis of any gene clustering algorithm is the similarity (or dissimilarity) measure between gene expression profiles. After every possible gene-pair has been considered, hierarchical clustering recursively pairs the genes in order of increasing dissimilarity. A new score is assigned to a gene group which is obtained by aggregating the scores of the genes within the group [2]. Thus, this process resembles bottom-up creation of a binary tree, resulting in a dendrogram as depicted in Figure 1. By applying a threshold, the tree is cut horizontally to form subtrees, each representing a cluster. Hierarchical clustering has the advantage of allowing a user to visualize how genes merge to form clusters. However, individual relationships between genes within a cluster is lost once the cluster is formed. For example, gene E in the example may be more similar to gene C, but not gene D. In this poster, we focus on the pairwise nature of the hierarchical clustering process and propose a method of microarray classification using gene pairs instead of single genes or gene clusters.