Figure 2

Cluster analysis on the combined dataset from U95Av2 and U133A. (a) The result of hierarchical clustering of 28 samples using the Pearson correlation coefficient as a distance measure. The dendrogram is exactly divided into the two groups representing U95Av2 and U133A, rather than by sample type. (b) The correlation coefficients between every two arrays in the combined dataset. The red (green) color corresponds to a higher (lower) value of correlation coefficient. The bias between U95Av2 and U133A is clearly represented here (upper left and lower right parts of the rectangle). (c) The result of clustering after gene scaling. Each gene expression value was scaled across the samples before combining U95Av2 and U133A datasets. In the dendrogram, the arrays obtained from the same biopsy are now joined together in all cases. (d) The bias between U95Av2 and U133A has clearly disappeared by gene scaling. The dark red diagonals in the upper right and lower left parts denote the high correlation coefficients for the same biopsy hybridized on different arrays.