Fig. 6 | Nature Communications

Fig. 6

From: Dissection of gene expression datasets into clinically relevant interaction signatures via high-dimensional correlation maximization

Fig. 6

Survival differences within DLBCL. a, b To visualize the survival trend revealed by our continuous 2-signature predictor, we show Kaplan-Meier estimates for partitions of predicted molecular risk (of ≥175%, ≥100%, ≥1⁄175%, and <1⁄175% the average risk in the detection cohort). Significant differences in progression-free survival were predicted in both the detection cohort (a, 470 R-CHOP treated DLBCL patients) and the validation cohort (b, 220 patients). (Statistics of each survival curve and OS can be found in Supplementary Fig. 30.) c, d Within GCB DLBCL (as originally classified), the quartile containing highest predicted molecular risks revealed significantly and strongly adverse survival, whereas the other three quartiles shared average survival (merged into one curve; details and ABC DLBCL are depicted in Supplementary Fig. 32; OS and second validation cohort in Supplementary Fig. 33). e, f To analyze and visualize the relationship between molecular risks and known clinical risks, we repartitioned all samples having an international prognostic index (IPI)42 into terciles. In each clinical risk class, significant survival differences remained between top and bottom terciles of predicted molecular risks (details in Supplementary Fig. 35).

Back to article page