CellMATE uniquely unveils cellular plasticity and nuanced spatial structure on joint profiling of transcriptome and chromatin accessibility. (A-B) Clustering performance of CellMATE and the competing methods evaluated by ACC (X axis) and ARI (Y axis) on scRNA+scATAC datasets profiled on sci-CAR kidney (A) and 10X PBMCs (B). (C) UMAP embeddings of the 10X PBMCs dataset learned by CellMATE, with CD14 monocytes highlighted (left). The inset shows the subgroups of CD14 monocytes uncovered by CellMATE (right). (D) Expression from scRNA (top) and gene activity from scATAC (bottom) of marker genes of stimulated monocytes in the two subgroups of CD14 monocytes. (E) Spatial distribution of all clusters learned by CellMATE or competing methods in spatial RNA + ATAC of mouse brain. Each dot is colored by cluster labels. (F) Main structures in the mouse brain. (G) Spatial distribution of cluster0 (C0) from CellMATE in spatial RNA + ATAC of mouse brain. (H) Spatial mapping accessibility of Unc5d, a marker gene of layer 4 of cortex, in spatial RNA + ATAC.