P02.03 Automated cell type specific PD-L1 quantification by artificial intelligence using high throughput bleach & stain 15-marker multiplex fluorescence immunohistochemistry in human cancers

Background The quantification of PD-L1 (programmed cell death ligand 1) has been used to predict patient’s survival, characterize the tumor immune microenvironment, and response checkpoint therapies. However, a framework assess status with high interobserver reproducibility on cells different types yet be established. Materials Methods To study impact expression microenvironment patient outcome, for fully automated was established validated. Automated facilitated by incorporating three deep learning steps analysis more than 80 neoplasms from 10’000 specimens using bleach & stain 15-marker multiplex fluorescence immunohistochemistry panel (i.e., PD-L1, PD-1, CTLA-4, panCK, CD68, CD163, CD11c, iNOS, CD3, CD8, CD4, FOXP3, CD20, Ki67, CD31). Clinicopathological parameter were available 30 entities overall survival data 1517 breast cancer specimens. Results Comparing deep-learning based conventional brightfield revealed concordance in (p<0.0001) as well an accuracy ranging 90% 95.2%. Across all entities, level significantly higher distinct macrophage/dendritic (DC) subsets (identified iNOS; p<000.1) macrophages/DCs located Stroma compared intratumoral macrophages/DC subsets. highly variable driven phenotypes identified intensity both cells, distance between non-exhausted T-cell (i.e. PD-1 CTLA-4 CD3 + CD8 cytotoxic T-cells, CD4 T-helper FOXP3 regulatory T-cells) macrophage/(DC) subtypes. In cancer, showed predictive performance area under receiver operating curves (AUC) 0.72 percentage (AUC: 0.54). positive negative cancers close spatial relationship T- (CD3 ± ) Macrophage/DC (CD68 CD163 CD11c iNOS) found prognostic relevant (p<0.0001). Conclusions conclusion, immunofluorescence assessment provides cutoff-free/continuous which are superior relevance. combined PD-L1/PD-1 20 subtypes phenotypes. Disclosure Information N.C. Blessin: None. E. Bady: T. Mandelkow: C. Yang: J. Raedler: R. Simon: Fraune: M. Lennartz: S. Minner: Burandt: D. Höflmayer: G. Sauter: S.A. Weidemann: