Fig. 3: Spatial cell-type signatures using PCF proteomics for resistance and response.
a, Univariable analysis of PFS at 2 years in relation to different cell types within the tumor compartment of the Yale cohort, using the median as a cut point. HRs and 95% CIs were derived from Cox proportional hazards models, and two-tailed log-rank P values were calculated and BH adjustment applied. Each data point represents one ROI from each patient. A total of n = 32 patients were analyzed using spatial proteomics. b, A representative ROI from a patient who progressed, characterized by a high presence of granulocytes and vascular structures. c, Univariable analysis of PFS at 2 years in the stromal compartment of the Yale cohort, highlighting the impact of different cell types on disease progression over the 2-year timeframe, with the median used as a cut point. d, The Kaplan–Meier plot shows the performance of the resistance signature predicting PFS at 5 years from the tumor compartment of the training cohort. e, Performance of the resistance signature predicting PFS at 2 years from the tumor compartment of the training cohort. f, Validation of the resistance signature predicting PFS at 2 years in the tumor compartment of the UQ validation cohort. g, Performance of the cell-type-response signature predicting PFS at 5 years from the stromal compartment of the training cohort. h, Performance of the response signature predicting PFS at 2 years from the stromal compartment of the training cohort. i, Validation of the response signature predicting PFS at 2 years in the stromal compartment of the UQ validation cohort. Two-tailed and one-tailed log-rank tests are used on the discovery and validation cohorts, respectively, where the direction of the effect in the latter is chosen to match the direction of the effect in the former. CI, confidence interval.
