Decoding, Modeling, and Reprogramming Cells at Scale in the Era of Digital Biology

Decoding the biological “languages” underlying genetic and cellular states remains a major challenge. Single-cell omics measurements are transforming our understanding of biology; however, they are expensive and destructive, posing challenges for monitoring live cells in tissues and humans over time. Although imaging is non-destructive, low-cost, and scalable, it can be difficult to interpret.

We aim to develop generalizable and scalable experimental and computational frameworks that utilize generative AI to bridge the gap between different data modalities in biology. For example, we have developed a series of technologies that enable the prediction of single-cell omics data from non-destructive imaging modalities, the reconstruction of molecular dynamics over time in live cells through imaging, and the generation of cellular and tissue images from single-cell gene expression profiles (a “DALL-E for biology”). These technologies enable fast and scalable querying and prediction of multi-omics information underlying mammalian and microbial systems using imaging, ideally in real time, in live cells.

By combining multi-modal live-cell and label-free chemical imaging with genetic and chemical perturbations, we have developed massively scalable screening systems to investigate the functions and behaviors of cells across space and time. Translating the different “languages” (data modalities) of biology can unify various views of cell and tissue biology, greatly reducing the need for multiple measurements, towards an ultimate goal of building virtual cells and digital simulators of multicellular systems.

Speaker: Jian Shu, Massachusetts General Hospital, Harvard Medical School, Massachusetts Institute of Technology