Fig. 6: Additional independent validation of methylome-based models.
From: Prediction of brain metastasis development with DNA methylation signatures
a, Application of the 5-year BM predictor to the external TCGA LUAD tissue dataset yielded methylome risk scores that predicted distant metastasis development after controlling for clinical variables with Cox proportional hazards modeling (n = 244). The box with the whiskers displays the HR and 95% CI. b, The clinical-methylomic BM nomogram was applied to the TCGA LUAD tissue samples for distant metastasis prediction; the time-dependent AUROC analysis using a bootstrap resampling approach showed nomogram utility in this external dataset (n = 244). The box with whiskers displays the AUROC and 95% CI. c, Application of the ensemble of plasma methylome-based BM-versus-others classifiers to the additional independent validation cohort of BM samples, showing their accurate identification as BM and not others (n = 31). In the box plots, the central bars correspond to the median; the upper and lower distribution quartiles are displayed using boxes; and the 1.5× IQR is represented by the whiskers. d, MDS plot of the external brain tumor and non-brain tumor control tissue cohorts using independent plasma-based features (DMRs) identified in pairwise comparisons between BM, glioma, CNSL and control samples, showing distinct clustering of BM samples from others. * denotes a significant P value.
