Skip to main content
. 2015 Sep 3;6(30):30306–30316. doi: 10.18632/oncotarget.5080

Figure 3. Ability of the model to identify groups of functionally related genes in the different sets.

Figure 3

A–C. Heat map representation of the correlations between (A) the protein expression values within the TCGA dataset, (B) the metagene expression within the TCGA dataset, (C) the metagene expression within the Fin-her dataset. Cells are coloured according to Pearson correlation coefficient values, with green indicating positive correlation and red negative correlation. D. Network representation of the metagenes. Each node represents the genes up- or down-regulated in the metagene. Edges show metagenes sharing a significantly high number of genes. The use of a network clustering algorithm showes the tendency of these metagenes to cluster together according to their ER-status. E–F. (E) ER-alpha and (F) GATA3 RPPA-based metagenes known to be associated with ER-positive tumours were able to predict pathological ER status in the Fin-her dataset. G–H. Correlation of the immune related metagenes (G) Lck and (H) Syk with the percentage of TILs in the Fin-her dataset.