Self-organizing maps in mining gene expression data

Abstract Modern DNA microarray technology provides means of measuring gene expression patterns of the whole genome of simple organisms at once. Exploratory analysis of these large-scale expression datasets is becoming vital to extracting functional information from the measurements. We demonstrate how self-organizing maps (SOM) can be applied to exploratory analysis of gene expression data from a yeast DNA microarray database in order to very rapidly find gene families with similar expression patterns. SOM not only enabled quickly selecting the gene families identified in previous work, but it facilitated identifying additional genes with similar expression patterns. Identifying new families of genes also appears to be possible as demonstrated by additional clusters of genes discovered from the data. Moreover, further insight into the primary pattern variations that discriminate between the families became explicit.