Table 2.

Analysis techniques for microarray expression profiling

Method	Type of analysis
Hierarchical agglomerative	Forms clusters starting at lowest level12
Hierarchical divisive	Forms clusters starting at the highest level14,15
Self organising maps	Partitions genes/experiments into prespecific number of groups by mapping onto nodes16,17
K-means	Partitions genes/experiments into prespecific number of groups by finding centroids7
Quality cluster algorithm	Forms groups based on diameter quality measure and jackknife correlations18
Super vector machines	Machine learning process that incorporates non-array information19
Singular value decomposition	Groups genes/experiments by reducing data matrix to eigen vectors (similar to principal component analysis)20,21
Gene shaving	Seeks groups of genes to maximise variation among experiments22
Class prediction	Identifies genes whose behaviour predicts defined classes23