Bridging single-cell genomics and tissue spatial organization in health and cancer

Abstract

The genomic profile of cells manifests itself in the local and global geometry of tissues. Tissue organization informs both normal physiological function in times of health, such as the rapid deployment of immune cells to stave off pathogens, and pathological dysfunction in times of disease, such as the uncontrolled proliferation of cancer cells. Understanding the connection between genomic information, such as single-cell gene expression levels, and spatial information, such as immune infiltration in tumor microenvironments, will aid in better diagnosing and treating patients. Herein, we explore the utility of cutting-edge mathematical approaches in elucidating this connection. In particular, we find that a principled, mechanistically-grounded mathematical approach engenders greater understanding of how cells of various types (e.g., immune, stromal, malignant) spatially interweave to facilitate cell-cell interactions and overarching spatial structures. This body of work charts the progressive incorporation of biological information, from black-box deep learning to interpretable machine learning to mechanistic mathematical modeling.