Fig. 3

Construction of prognostic risk prediction models. The samples of the A training set, B validation dataset 1 (GSE27020), and C validation dataset 2 (GSE25727) are all based on the KM curve of the PS prediction models and prognosis, with blue and red curves representing low- and high-risk samples, respectively (D). The ROC curve of the prediction results was based on the PS prognosis model