Deep learning to integrate histology with spatial transcriptomics

I will present our new computer vision algorithm, ST-Net, which can computationally synthesize spatially resolved transcriptomics directly from H&E histology images (He et al. Nature Biomedical Engineering 2020). I will demonstrate ST-Net on breast cancer data, and show how it characterizes tumor spatial heterogeneity and quantifies tumor-immune interactions. I will then discuss how similar algorithms can capture complex morphological changes over time of individual cells.

Learning Objectives:

1. Understand how deep learning can integrate imaging with genomics

2. Understand how to characterize spatial and temporal variation in cells and connect it with human diseases