Research Paper Volume 13, Issue 13 pp 17592—17606

DNA methylation biomarkers for diagnosis of primary liver cancer and distinguishing hepatocellular carcinoma from intrahepatic cholangiocarcinoma

Screening of diagnostic sites. The top 10 methylation sites with indicated Gini values for distinguishing PLC and normal (A). The AUC curves of the diagnostic prediction model (PLC versus Normal) in the training and validation sets based on indicated sites combination (B). Evaluation of diagnostic model (PLC versus Normal) based on three sites (C). The receiver operating characteristic (ROC) curves of diagnostic model (PLC versus Normal) based on three sites in the training (D) and validation sets (E). The top 10 methylation sites with indicated Gini values for distinguishing HCC and ICC (F). The AUC curves of the diagnostic prediction model (HCC versus ICC) in the training and validation sets based on indicated sites combination (G). Evaluation of diagnostic model (HCC versus ICC) based on three sites (H). The receiver operating characteristic (ROC) curves of diagnostic model (HCC versus ICC) based on three sites in the training (I) and validation sets (J).

Figure 2. Screening of diagnostic sites. The top 10 methylation sites with indicated Gini values for distinguishing PLC and normal (A). The AUC curves of the diagnostic prediction model (PLC versus Normal) in the training and validation sets based on indicated sites combination (B). Evaluation of diagnostic model (PLC versus Normal) based on three sites (C). The receiver operating characteristic (ROC) curves of diagnostic model (PLC versus Normal) based on three sites in the training (D) and validation sets (E). The top 10 methylation sites with indicated Gini values for distinguishing HCC and ICC (F). The AUC curves of the diagnostic prediction model (HCC versus ICC) in the training and validation sets based on indicated sites combination (G). Evaluation of diagnostic model (HCC versus ICC) based on three sites (H). The receiver operating characteristic (ROC) curves of diagnostic model (HCC versus ICC) based on three sites in the training (I) and validation sets (J).