Fig. 4: Validation of ICBcircSig score model in two independent cohort.

a Boxplot of ICBcircSig score between response (n = 38) and non-response (n = 30) groups. b Receiver Operating Characteristic (ROC) curves quantifying ICBcircSig score prediction AUC. c Kaplan–Meier survival curves of PFS between high- and low-ICBcircSig score patients stratified by the optimal cutoff in cohort 2. d Time-dependent ROC curve at 12 and 24-months of PFS for the ICBcircSig score. e Forest plot for the HRs of multivariate Cox model of the ICBcircSig score and clinicopathological variables. f Kaplan–Meier survival curves of OS between high- and low-risk patients stratified by ICBcircSig score using the optimal cutoff in validation data. g Time-dependent ROC curve at 12 and 24-months of OS for the ICBcircSig score. h Forest plot for the HRs of multivariate Cox model of the ICBcircSig score for OS and clinicopathological variables. I, j Boxplot of expression of circTMTC3/circFAM117B (i) ICBcircSig score (j) distribution between response (n = 14) and non-response (n = 7) groups for in-house cohort 3. A log-rank test was used in c, f and i. k ROC curves quantifying ICBcircSig score response prediction AUC in in-house cohort 3. l Kaplan–Meier survival curves of PFS between high- and low-risk patients stratified by ICBcircSig score using the optimal cutoff in in-house cohort 3. m Time-dependent ROC curve at 12 and 24-months of PFS for the ICBcircSig score in in-house cohort 3. A two-sided Wilcoxon rank-sum test was used in a, i, and j. The boxes in a, i and j indicate the median ± 1 quartile, with the whiskers extending from the hinge to the smallest or largest value within 1.5× IQR from the box boundaries. Black vertical lines in e and h indicate the 95% confidence interval (CI). PFS progressive free survival, OS overall survival, ROC receiver operating characteristic curve, HR hazard ratio, AUC Area Under the ROC Curve. Source data are provided as a Source data Fig. 4a–d, f, g, i–l.