Figure 2

Clustering associated with ER-α expression in the NKI2, Wang and TRANSBIG data sets. All graphs were based on unsupervised hierarchical clustering analyses using the 5,000 randomly-generated unigene gene lists. A. Percentage of lists that segregate tumors into groups that contain a disproportionately high number of ER negative tumors as compared to ER positive tumors for each data set. A ratio greater than 1 indicates that the clusters contain a disproportionate number of ER negative tumors, as compared to ER positive tumors. The grey bars and black bars show the percentage of gene lists where this ratio is 2 and greater than 3, respectively. B. Percentage of lists that segregate tumors into groups that contain a disproportionately high number of ER negative tumors as compared to ER positive tumors in data sets that were globally adjusted for ER-α expression. C. Percentage of lists that were significantly predictive of metastatic recurrence latencies in unadjusted data sets (grey bars) or data sets that were globally adjusted for either ER-α expression (black bars) or a marker of proliferation (white bars). D. Percentage of lists that segregate tumors into groups that contain a disproportionately high number of ER negative tumors as compared to ER positive tumors in data sets that have been globally adjusted for a marker of proliferation.