Fig. 6: Cancer detection across cancer types.
From: Genome-wide mutational signatures in low-coverage whole genome sequencing of cell-free DNA
Receiver Operating Characteristic (ROC) curve for sample classification from the DELFI cohort, which included 199 individuals with cancer and 206 healthy individuals. A random forest model using ten-fold nested cross-validation with 500 iterations was used to classify samples as either healthy or cancer. This cohort consisted of stage I–IV NSCLC (n = 27); stage I–III breast cancer (n = 48); stage I–IV CRC (n = 27); stages I–IV, 0 and X gastric cancer (n = 27); stage I–IV ovarian cancer (n = 26); and stage I–III pancreatic cancer (n = 34). All ROC curves from 500 iterations are shown. Classification performance by individual cancer type and stage is shown in Supplementary Fig. 15. Source data are provided as a Source data file. AUC area under the curve; CI confidence interval; CV cross validation.
