Supplementary Figure 6: ROC analysis to measure the predictive power of CD38⁺ fraction of PD1⁺CD8⁺ T-cells pre- and post- anti-PD-1 therapy in the human tumor and PBMC samples.

a-c. The ROC curves were generated using R version 3.5.1 software in post-therapy tumors with 4% cut-off (a), pre-therapy tumors with 10% cut-off (b), and PBMCs with 5% cut-off (c). AUC and 95% confidence interval (CI) were determined using Delong method with R version 3.5.1 statistical software. The diagnostic tables used for generating each ROC curve as well as the formulae used to calculate sensitivity, specificity, PPV and NPV are provided. Comparison of responding vs. non-responding tumor lesions that had more than 4%^# or 10%^{^} PD-1⁺CD38⁺ cells in the CD8⁺ population in the TME. *For human PBMC data, the CD38⁺ fraction of PD-1⁺CD8⁺ T cells that showed more than 5% decline at 9 weeks when compared with 3 weeks were compared between responders and non-responders post-therapy. AUC: Area Under the Receiver Operating Characteristics (ROC) Curve; AUC 95% CI: There is 95% of confidence that the interval contains the true AUC. For example, there is 95% confidence that (0.679,1) contains the true value of AUC for 5% cut-off; PPP: Positive predictive value; NPV: Negative predictive value.