Abstract
Figure S1a: Representative images of the BLEACH&STAIN 15+1 staining. Figure S1b: BLEACH&STAIN 15+1 validation. Figure S2: Deep learning framework for automated cell type identification. Figure S3: Performance of a Deep Learning-based framework for cell type identification compared to manual set thresholds. Figure S4: Comparison between BLEACH&STAIN and manual PD-L1 assessment. Figure S5: Immune phenotype identification. Figure S6: Association between the immune phenotypes and the individual immune cell parameters. Figure S7: Representative image of the 6 major immune phenotypes of Figure 2 Figure S8a: Association between PD-L1 phenotypes and clinico-histopathological parameters. Figure S8b: Distribution of six PD-L1 phenotypes across 44 carcinoma entities. Figure S9: Spatial orchestration of immune cells across the PD-L1 immune phenotypes. Figure S10: Representative image of the spatial orchestration shown in Figure 3 Figure S11: Composition of immune cells across the three PD-L1 phenotypes Figure S12: Prognosis analysis of spatial immune parameters in breast cancer for overall 5-years survival after surgery. Table S1: TMA cohort of the PD-L1 15+1 study. Table S2: List of the used antibodies, antigen retrieval (AR), dilutions, and Opal dyes for multiplex fluorescence immunohistochemistry. Table S3: List of 132 immune cell (sub)population. Table S4: Associations between spatial parameters and PD-L1 phenotypes were independent from the used threshold. Table S5: Immune landscape of 3098 human carcinoma samples.
Keywords