Nuclear Receptor Expression Defines a Set of Prognostic Biomarkers for Lung Cancer
Figure 3
Kaplan-Meier plots showing the predictive power of the NR gene signature in datasets from the NCI Director's Consortium.
(A) LOOCV of the recursive partitioning tree model (RPART) for the 30-sample MDACC QPCR dataset using all 48 NRs. The HR for the predicted high-risk versus the predicted low-risk signatures was 13.6; 95% CI, 3.07–60.92; p = 0.000014. (B and C) Independent validation of the 48-NR gene expression signature between the MDACC cohort and the Consortium cohort. The MDACC cohort training set (n = 30) was tested in the Consortium cohort (n = 442) (B), and vice versa (C). (D) Independent validation of the NR gene signature in the 442-sample multi-institute Consortium using RPART analysis. The microarray datasets were divided into two groups, one for the training cohort (n = 256) and the other for the testing cohort (n = 186). p-values were obtained by the log-rank test. Red and black lines represent predicted high- and low-risk groups, respectively. Circles indicate censored samples. Note that SHP expression is the single co-variable (or predictor) in the classification model that describes the survival time differences shown in (B), demonstrating that SHP is a single gene predictor that represents the entire 48-NR gene profile.