Fig. 4
From: Identification of prostate cancer specific methylation biomarkers from a multi-cancer analysis
Model construction results of prostate cancer specific biomarkers. a AUC distribution of models in separating Gleason 6 and Gleason 3 + 4 prostate tumors. Models built by 8 biomarkers + PSA + age resulted in AUCs 0.57 ± 0.005. Models built by 8 biomarkers resulted in AUCs 0.63 ± 0.004. Models built only by PSA + age resulted in AUCs 0.46 ± 0.005. b AUC distribution of models in separating Gleason 6 with Gleason 4 + 3 and 8–10 prostate tumors. Models built by 8 biomarkers + PSA + age resulted in AUCs 0.84 ± 0.003. Models built by 8 biomarkers resulted in AUCs 0.87 ± 0.002. Models built only by PSA + age resulted in AUCs 0.67 ± 0.004. c ROC of three validating GEO datasets in separating normal tissue and prostate tumors: GSE47915 in black (AUC = 0.88), GSE76938 in blue (AUC = 0.89), GSE112047 in red (AUC = 0.92). D) ROC of one GEO dataset in separating prostate tumors and other urinary related samples: GSE52955 (AUC = 1)