Fig. 7

Clustering cell types in the MERFISH dataset. A Barplots of ARI display the accuracy of different methods for cell type clustering. Compared methods include BASS, Seurat, SC3, and FICT. Results for BASS, Seurat, and SC3 are shown for both the single tissue section analysis on the tissue section Bregma-0.14 and multi-sample analysis (BASSMult, SC3Mult, and SeuratMult) that fitted all five tissue sections (Bregma-0.04, -0.09, -0.14, -0.19, and -0.24) and evaluated ARI on the section Bregma-0.14 (e.g., BASSMult (Bregma-0.14)) as well as on all five tissue sections (e.g., BASSMult). B–E UMAP visualization of cell type clustering results on Bregma-0.14 by B BASS, C Seurat, D SC3, and E FICT for analyzing the single section Bregma-0.14. F UMAP visualization of cell type clustering results on five sections by the multi-sample version of BASS. G Cell type compositions estimated by the multi-sample version of BASS. Excitatory neurons: E1, E2, E3, E4, and E5; Inhibitory neurons: I1, I2, I3, I4, I5, I6, I7, and I8; MOD: mature oligodendrocytes; IOD: immature oligodendrocytes; Astro: astrocytes; Micro: microglial cells; Epen: ependymal cells; Endo: endothelial cells; and Mural: mural cells