AstraZeneca: Computational Pathology: Clearing the way for more robust IHC biomarker diagnostics

AI-enabled computational pathology is transforming cancer therapeutic target assessment, delivering precision, accuracy, and reproducibility that far exceed visual scoring by pathologists. By detecting and quantifying target expression heterogeneity, new biological insights can be derived to enable novel biomarkers for patient selection and therapeutic response prediction. Over the past 6+ years, AstraZeneca has developed Quantitative Continue Scoring (QCS) technology, a supervised AI-based computational pathology solution. Trained on expert pathologist-annotated digital images, QCS algorithm can detect and segment tumor regions, tumor cells and their subcellular components automatically, and subsequently quantify therapeutic target expression on cell membrane and cytoplasm. Notably, based on pre-specified human-interpretable features, QCS algorithm can derive additional biological characteristics, including target internalization and target heterogeneity, which can be utilized for the prediction of therapy response to antibody-drug conjugates (ADCs), given that the quantification of the target expression on the membrane and cytoplasm of tumor cells may provide insights into target internalization. These AI-based advancements in computational pathology hold the promise of transforming diagnostic pathology and unlocking new frontiers in companion diagnostic solutions, thereby maximizing the potential of personalized oncology with unprecedented precision.