Fig 4.

Prediction performance of PreMSIm in predicting MSI. A, ROC curve analysis of TCGA colon cancer. B, ROC curve analysis of pan-cancer. All pan-cancer samples were separated into training (80% of samples) and test sets (20% of samples). In the training set, the 10-fold CV AUC was shown. C, ROC curve analysis of TCGA gastric and colon cancers using the TCGA endometrial cancers as the training set. D and E, ROC curve analysis of two gastric (D) and two colorectal (E) cancer cohorts in which the PreMSIm R package was used to predict MSI. MSI: microsatellite instability. CV: cross validation. AUC: area under the ROC curve. COAD: colon adenocarcinoma. STAD: stomach adenocarcinoma.