Deep Learning Predicts Subtype Heterogeneity and Outcomes in Luminal A Breast Cancer Using Routinely Stained Whole-Slide Images.

abstractIntratumor heterogeneity (ITH) presents challenges for precision oncology, but methods for its spatial quantification, scalable at population levels, do not exist. Based on previous work showing that the admixture of PAM50 subtype can be measured from bulk tissue using transcriptomic data, we trained a deep convolutional neural network to quantify subtype ITH in luminal A (LumA) breast cancer from routinely stained whole-slide images. We tested the hypothesis that subtype admixture detected in images was associated with tumor aggressiveness and adverse outcomes. In 680 cases from the TCGA-BRCA cohort, we determined adherence to assigned subtype by applying matrix factorization to each transcriptome. The purest cases were split into groups for initial testing, training, and parameter tuning. A total of 230 LumA-assigned cases were held out for final testing. Image patches were fed into a deep neural network pretrained on histology images. We measured the association of tumor area classified as LumA in the image with tumor characteristics and survival. Among LumA-assigned cases, admixture was associated with slightly higher estrogen receptor positivity but lower progesterone receptor (PR) positivity and estrogen receptor–related gene expression and higher HER2 positivity, tumor size, grade, and tumor–node–metastasis stage. Image admixture was associated with more TP53 and fewer PIK3CA mutations. Progression-free survival was significantly shorter in more admixed cases. Our findings demonstrate that deep learning, trained to recognize genomic correlates in tissue morphology, can quantify and map subtype admixture in LumA breast cancer that has clinical significance. The low cost and scalability of this method hold potential as a research tool for investigating ITH and perhaps improving the efficacy of precision oncology. SIGNIFICANCE: A deep learning model, trained using transcriptomic data, inexpensively quantifies and fine-maps ITH due to subtype admixture in routine images of LumA breast cancer, the most favorable subtype. This new approach could facilitate exploration of the mechanisms behind such heterogeneity and its impact on selection of therapy for individual patients.