Fig. 4

Integration of scRNA-seq data of cerebral organoids with fixed and unfixed experimental conditions. a Schematic of the experimental design. Fresh and methanol-fixed single-cell suspensions of the same four-month-old cerebral organoid were measured using scRNA-seq. b–d UMAP embedding of the scRNA-seq data before and after integration with the seven integration methods. Cells are colored by b experimental conditions, c RSS-based cell type annotations, and d model-based predicted cell type annotations. e, f Proportion of neighbors of each cell in different categories, with neighbors defined as 50 cells with the shortest Euclidean distances with a cell in PCA (no integration) or different integration spaces. Cell types are annotated based on e clustering on RSS or f model-based prediction. g LISI-based scores of conditions (left), RSS-based cell type annotation (middle), and model-based cell type annotation (right) with no or different integration methods. h Heatmaps showing summarized performance metrics. Numbers show the ranking of different integration methods based on different metrics