Extended Data Fig. 5: Tumor tissue IHC analyses and correlation with tracer signal.

a, Representative examples of IHC CD8 expression phenotypes (n = 34): i, Liver biopsy of a cholangiocarcinoma metastasis with a desert phenotype. ii, A biopsy of a perivesical tumor mass of dMMR urothelial cell cancer with stromal CD8 expression phenotype [density 2]. iii, A liver biopsy of dMMR colon carcinoma shows an inflamed phenotype [density 3]. b, Correlation of mean CD8 staining pixel positivity and autoradiography signal across 16 samples with weighted quantile regression fit. Point sizes and regression weights are proportional to the size of each sample biopsy. c, Using tile-based analysis, the correlation across and within samples of mean CD8 staining pixel positivity and autoradiography signal at subsample level. Cross-sample correlations and corresponding 95% confidence intervals are displayed with horizontal and vertical lines at each tile size. Within-sample correlations are presented at each tile size for each sample as circles. At each tile size, only samples with ≥ 6 tiles are shown. Tiles containing less than < 25% tissue were excluded. Only cross-sample correlations are shown at tiles sizes higher than 5000px2 as no single sample had >5 tiles. d, Tiles of varying sizes are shown for a single representative clear cell ovarian cancer sample. e, Violin plot of tumor SUV_max-to-muscle SUV_mean ratio with desert (n = 15) and non-desert (n = 19) immune phenotype before and during treatment in 24 patients. f, Violin plot of SUV_max in lesions with desert (n = 15), stromal (n = 15) and inflamed (n = 4) immune phenotype before and during treatment in 24 patients. e-f, Coloured dots show individual lesions; black vertical lines show 95% CI of the geometric mean; white dots within black lines and values below the violin plot the actual geometric means. Two-sided nominal P-values were derived from linear mixed models taking clustering within patients into account, either using a likelihood ratio test under maximum likelihood (e; P_trend in f) or using a Wald test under restricted maximum likelihood (P-values for factor levels in f).