scEGOT: single-cell trajectory inference framework by entropic Gaussian mixture optimal transport

Time-series single-cell RNA sequencing (scRNA-seq) data in biology has opened the door to elucidate cell differentiation processes. In this context, the optimal transport theory has attracted attention to interpolate scRNA-seq data in adjacent times and infer the trajec- tories of cell differentiation. This paper presents scEGOT, a novel comprehensive single-cell trajectory inference framework based on entropic Gaussian mixture optimal transport (EGOT). The method- ology facilitates the inference of cell differentiation pathways and dynamics from time-series scRNA-seq data. The scEGOT frame- work provides comprehensive outputs, including cell state graphs, velocity fields of cell differentiation, time interpolations of scRNA- seq data, space-time continuous videos of cell differentiation with gene expressions, gene regulatory networks, and reconstructions of Waddington's epigenetic landscape. These outputs allow us to un- derstand the dynamics of the cell differentiation process from mul- tiple perspectives. To demonstrate that scEGOT is a powerful and versatile tool for single-cell biology, it was applied to time-series scRNA-seq data of the human primordial germ cell-like cell (human PGCLC) induction system. This application provided new insights into the mechanism of PGCLC/somatic cell segregation. In particu- lar, we identified the PGCLC progenitor population and the time point of the segregation. In addition, we found novel key genes that may be critical for human PGCLC differentiation. Overall design: Time-series single-cell RNA sequencing (scRNA-seq) data. For the aggregates of PGCLC induction at day 0.5, 1, 1.5 and 2, the aggregates were collected.