[SS21] Transformation and Integration of Data and Modeling in Medicine

[SS21-06] Inferring Cell Differentiation Dynamics from Time-Series scRNA-seq Data Using scEGOT

*Toshiaki Yachimura¹ (1. Tohoku University (Japan))

Keywords:

Trajectory inference,Optimal transport,Gaussian mixture model,Epigenetic landscape,Single-cell biology

The epigenetic landscape, proposed by C.H. Waddington in 1957, is a conceptual model that visually represents how cells follow different fates during the process of differentiation. In recent years, with the development of measurement technologies such as single-cell RNA sequencing (scRNA-seq), there has been growing interest in reconstructing this conceptual model based on gene expression data and inferring the trajectories of cell differentiation. In this talk, I will present scEGOT, a novel comprehensive trajectory inference framework for cell differentiation using time-series scRNA-seq data, based on entropic Gaussian mixture optimal transport (EGOT). scEGOT allows us to infer not only the cell state graph of cell differentiation constructed by conventional trajectory inference methods, but also the velocity field and dynamics of gene expression. Furthermore, our method reconstructs Waddington's epigenetic landscape in gene expression space and infers gene regulatory networks. Additionally, I will discuss the application of scEGOT to time-series scRNA-seq data during the induction of human primordial germ cell-like cells (hPGCLCs) from iPS cells. I will show our results in identifying the progenitors of hPGCLCs and the marker genes involved in their induction.