MicroCluster is a deterministic biclustering algorithm that can mine arbitrarily positioned and overlapping clusters of gene expression data to find interesting patterns. Depending on the parameter values, MicroCluster can mine different types of clusters, including those with constant or similar row or column values, as well as scaling and shifting expression patterns. MicroCluster first constructs a range multigraph, a compact representation of all value ranges in the data set that are similar between any two columns. It then searches for constrained maximal cliques in this multigraph to yield the final set of biclusters. Optionally, MicroCluster merges or deletes clusters with large overlaps. Tests on several synthetic and real data sets illustrate MicroCluster's effectiveness.

This article is part of a special issue on data mining for bioinformatics.