Introduction

CLICK (CLuster Identification via Connectivity Kernels) is a newer algorithm for clustering [20]. The input for CLICK is the gene expression matrix. Each row of this matrix is an ``expression fingerprint'' for a single gene. The columns are specific conditions under which gene expression is measured (e.g. different points in time). A more formal definition is as follows: Let $N=\{e_1,\ldots,e_n\}$ be a set of elements. Let M be an input real-valued matrix of order $n\times p$, where M_ij is the j-th attribute of e_i. The i-th row-vector in M is the fingerprint of e_j. For a set of elements $K\subseteq N$, we define the fingerprint of K to be the mean vector of the fingerprints of the members of K. One seeks to partition N into clusters (subsets). In such a partition, elements in the same cluster are called mates. The CLICK algorithm attempts to find a partition of N into clusters, so that two criteria are satisfied: Homogeneity - mates are highly similar to each other; and separation - non-mates have low similarity to each other.