Figure 3

Association patterns between copy number and gene expression found on 17q for two independent breast cancer studies. Heatmap of association structure (green, positive; red, negative; black, no association) for the datasets of Pollack (A) and Chin (B), with rows representing copy number probes from centromere (bottom) to telomere (top), and columns representing gene expression probes from centromere (left) to telomere (right). The vertical bars on the left-hand side represent the p-values for the copy number probes, as calculated by the gene-set model (bar I) and by the window gene-to-gene model (bar II), with blue indicating the significant ones (for Pollack FDR ≤ 0.10, for Chin FDR ≤ 0.01). The top horizontal bar indicates expression probes with strong positive (green) or negative (red) association with the significant results from the gene-set model (mean z-score across significant tests ≥ 3). In the left panel, the log-ratio copy number values for all samples are represented by their smoothed medians, on the same probe spacing (equal space between each pair of consecutive probes) as used in the heatmap for comparability.