Extended Data Fig. 10: CellOracle noto LOF simulation with Farrell et al. zebrafish embryogenesis data.
From: Dissecting cell identity via network inference and in silico gene perturbation
(a) 2D force-directed graph of a published atlas32 of zebrafish embryogenesis (n = 25,711 cells). (b) Main trajectory partitioned into four sub-branches. (c) Bar plots depicting the number of TFs after variable gene selection (black), the number of TFs with >1 network edge in the inferred GRN model (dark grey), and the number of TFs expressed in >1% of cells (light grey). (d) CellOracle noto LOF simulation result (left) and simulation results with a randomized GRN model (right) for the notochord lineage. Simulated cell state transitions for each cell were converted to a vector field and visualized with a scatter plot (shown in grey). (e) Noto LOF simulation for the prechordal plate lineage. (f) CellOracle noto LOF simulation vector is shown at single-cell resolution. Cells in the Notochord cluster are shown in orange, while the Prechordal Plate cells are shown in blue. The right panel is the magnified area. (g) Force-directed graph of the Other mesendoderm sub-branch with cell cluster annotations from the Farrell et al. study32 (n = 10,265 cells). (h) Pseudotime data are projected onto the force-directed graph. (i) The Somite lineage, defined in the previous Farrell et al. study32, is in red. (j) Pseudotime gradient vector field calculated for the Somite lineage. (k) Noto LOF simulation vector field in the cells of the Somite lineage are shown with perturbation scores.
