Phenotypic composition of salivary gland tumors: an application of principal [corrected] component analysis to tissue microarray data.

The tissue organization of the salivary gland is complex, and a large number of salivary gland tumor entities with a broad morphologic spectrum are listed, creating tumor classification schema for the salivary glands that are difficult to understand. In the present study, we attempted to examine how the anatomical components of the salivary gland are associated with morphological subtypes of tumors. We selected a panel of 12 molecules, which labeled one or some of the components, with all of the markers covering every component of the salivary glands. Using tissue microarray, expression profiles of these molecules were examined in four representative spots from each of 88 salivary gland tumors. The resulting large data matrix was analyzed using principle component analysis (PCA). We considered the first three eigenvectors to be significant; as the eigenvalues were more than 1.0 and the cumulative proportion achieved was 67%. Comparison with expression patterns in normal tissue suggested that the three components represented myoepithelial differentiation, and luminal and basal cell phenotypes. Then, we compared the PCA results with individual morphologic subtypes. Individual subtypes were clustered among the three dimensions of the components. This implies that salivary gland tumors may be well characterized by using only three components. Copyright 2004 USCAP, Inc.