ICI: a new approach to explore between-cluster relationships with applications to gene expression data

Clustering methods are typically used to group objects with similar patterns into one of k sets. In the case of gene expression data, genes clustered into the same set tend to have similar expression profiles. However, clustering techniques lack a mechanism to quantify the between-cluster relationship. We propose a method called ICI (Inter-Cluster Investigator) to ascertain the between-cluster and between-gene relationships and establish the amount of co-regulation between clustered sets of genes. This will yield an alternative way of characterizing the between-gene relationship that builds upon the existing structure obtained from clustering. A procedure is proposed to allow the identification of negatively and positively correlated sets of genes, which may indicate repression or regulation in gene expression data. It is illustrated with the analysis of yeast cell-cycle data.