Contrastive learning-based histopathological features infer molecular subtypes and clinical outcomes of breast cancer from unannotated whole slide images

Hui Liu, Yang Zhang, Judong Luo

Abstract: Highlights•We verify contrastive learning extracts more informative pathological features.•Our model successfully infers the expression levels of tumor recurrence-related genes.•Our method effectively establishes the high-order genotype-phenotype associations.•ST-seq data verified model ability in constructing gene spatial expression landscape.