Please join the Goergen Institute for Data Science for Using Data Science to Uncover the Morphogenetic Blueprints of the Fly Embryo, a research seminar with Tomer Stern, Postdoctoral Research Fellow at Princeton University.

Bio: Tomer received his Ph.D. in Computational Biology at the Weizmann Institute of Science, Israel. There, together with Eli Zelzer, he focused on the mechanisms underlying early morphogenesis of long bones and cartilage in the mouse model, through computer vision analysis of micro-CT images.

In 2016, Tomer began his postdoctoral training at Princeton University, together with Eric Wieschaus and Stas Shvartsman. One direction of his research focuses on generation of the first whole embryo mapping of morphogenetic domains during gastrulation of the fruit fly, using advanced computer vision and time-series data mining. His second direction is dedicated to the application of graph neural networks for learning tissue dynamics directly from in-silico data, as the basis for data-driven modeling of tissue morphogenesis.

He was awarded multiple fellowships, including the Hammer Prize of the Israeli Ministry of Education, The Otto Schwartz Prize for academic excellence, and the EMBO long term fellowship award.

Abstract: The development of multicellular organisms is carried through an intricately orchestrated interplay of various cell behaviors, including division, delamination, shape changes, and neighbor exchanges. Using advanced computer vision and time-series data mining, we have developed a template-matching algorithm for searching dynamic cell behaviors in live imaging data. With this approach we generated the first whole embryo single cell atlas of morphogenetic activity underlying gastrulation in the fruit fly. It has allowed us to identify and characterize differences between mitotic domains and has revealed a previously unknown large dorsal intercalary domain. This work opens new possibilities in studying the early design principles at the whole organism level.

In the second part of the talk I will present “Deep Cell Graphs” (DCG), a graph neural network designed for learning the intricate geometric and mechanical interactions between cells of a computer simulated epithelial tissue. DCG was able to accurately predict present mechanical properties as well as future topological changes in the tissue, thereby providing a significant step toward data-driven modeling of large cellular network interactions in tissue morphogenesis.