Development of an Independent Prognostic Signature Based on Three Hypoxia-Related Genes for Breast Cancer

Figure 4

The prognostic model could efficiently predict the survival of BC patients. (a–c) The Risk Score distribution of samples from TCGA database and GEO cohort. One point represented a sample; the red points were the samples with high Risk Score; the green points were the samples with low Risk Score, and the intersection point was the optimal Risk Score. (a) Training set from TCGA database, (b) Testing set from TCGA database, and (c) GSE42568 and GSE48391 cohorts. (d–f) The heat map of mRNA expression values of three HRGs in samples from TCGA database and GEO cohort. The horizontal axis represented genes; the vertical axis represented samples; red indicated high expression; blue indicated low expression. The categories of samples were marked with different colors on the top of the heat map. (d) Training set from TCGA database, (e) Testing set from TCGA database, and (f) GSE42568 and GSE48391 cohorts. (g–i) The Kaplan Meier survival curve of samples from TCGA database and GEO cohort. (g) Training set from TCGA database, (h) Testing set from TCGA database, and (i) GSE42568 and GSE48391 cohorts. (j–l) The time-dependent ROC curve of samples from TCGA database and GEO cohort. The horizontal axis was False Positive; the vertical axis was True Positive; the accuracy of prediction was evaluated by AUC value (the area under the ROC curve). (j) Training set from TCGA database, (k) Testing set from TCGA database, and (l) GSE42568 and GSE48391 cohorts.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)

(l)