Mor Nitzan, PhD, is a John Harvard Distinguished Science Fellow and James S. McDonnell Fellow at the Paulson School of Engineering and Applied Sciences at Harvard. She will deliver the talk, "Geometric Embeddings for Single Cell Data" as part of the PMG Seminar Series. 

Abstract: Massively multiplexed sequencing of RNA in individual cells is transforming basic and clinical life sciences. In standard experiments, however, tissues must be first dissociated. Thus, crucial information about spatial relationships between cells, along with the tissue-wide expression patterns they confer, is lost. This poses a fundamental problem for elucidating collective function of tissues, developmental pathways, and mechanisms of cell-to-cell communication. Considerable efforts to overcome this challenge have been undertaken. However, experimental methods are either technically challenging, or have limited resolution or throughput. Existing computational approaches predict spatial positions by comparing each sequenced cell, independently, to an imaging-derived spatial gene expression database for that tissue. However, these approaches rely on prior knowledge of spatial expression patterns which often does not exist, or is difficult to construct. Here, we explore a different idea. We postulate that cells in spatial proximity, overall, share more similar transcriptional profiles than cells farther apart. We validate this hypothesis for several complex biological systems. Consequently, we seek to find spatial arrangements of sequenced cells on tissue space which optimally preserve this principle. We show that this hard optimization problem can be cast as a generalized optimal transport problem for probabilistic embedding, for which we derived an efficient iterative algorithm. We successfully reconstruct the mammalian liver and intestinal epithelium, as well as fly and zebrafish embryos. Our results demonstrate a simple spatial expression organization principle which can be used to infer meaningful spatial position probabilities for individual cells. Our framework (“novoSpaRc”) is flexible, can naturally incorporate prior spatial information, is scalable to large number of cells and compatible with any single-cell technology. I will further discuss how these ideas can be extended to incorporate generalized principles underlying spatial organization of gene expression for more complex tissues, and be employed for the inference of a variety of biological processes from single cell data.