We consider the model trained based on the top N (N = 1,351 for AML; N = 2,137 for BRC) DISCERN-scoring genes and clinical covariates (blue), and the model trained based on only clinical covariates (red) (panels A and C for AML3 and BRC3, respectively). (B) The panel shows the comparison with the model trained using genes comprising 22 genes previously known prognostic marker, called LSC [54], along with the clinical covariates (red). (D) The panel shows the comparison with the model trained using 67 genes from the MammaPrint prognostic marker (70 genes) [83] along with the clinical covariates. We used 67 genes out of 70 genes that are present in our BRC expression datasets. P-values shown in each plot are based on the logrank test (red).