Fig. 3

CSS integrates time-course scRNA-seq data of cerebral organoid development from pluripotent stem cells. a Schematic of the experimental design. PSC, pluripotent stem cell; EB, embryoid body; NEcto, neuroectoderm; NEpith, neuroepithelium. b Joint analysis of the time course cerebral organoid scRNA-seq data without integration. Dots, each representing one cell, are colored by sample time points (left) and PSC lines (right), respectively. Expression patterns of example markers are shown as feature plots. Org, organoid; m, month. c–h Joint analysis of time course scRNA-seq data with c CSS, d Scanorama, e MNN, f Harmony, g Seurat v3, and h Harmony integration. The cells in the UMAP embedding are colored by sample time points (left) and PSC lines (right), respectively. i Stacked bar plots showing average proportions of cells from different time points in the neighborhood of cells in different time points when no integration or different integration methods are used. j Proportion of PSC neighbors in different annotation categories. PSC is the union of cells in PSC and EB time points. Neighbors are defined for each cell as top 50 cells with the shortest Euclidean distance in PCA (no integration) or other integrated representation spaces. k Average Euclidean distances from PSC to cells at later time points in PCA and different integrated representation spaces (bottom). l Intermediate cells between PSC and neuroectoderm (NEcto) stages are defined as those with comparable proportion of PSC and NEcto neighbors in the CSS space (top). RNA velocity analysis (bottom left) supports a potential cell state transition from PSC (dark blue) to NEcto (light blue) via the identified intermediate state (yellow cells). Expression patterns of two genes marking the intermediate state, CYP26A1 and LITAF, are shown (bottom right)