Fig. 2
From: Unsupervised spatially embedded deep representation of spatial transcriptomics
Quantitative assessment of SEDR on the human dorsolateral prefrontal cortex (DLPFC) dataset. A Manual annotation for the DLPFC #151673 section and clustering results of eleven methods (SpatialLDA, Seurat, Giotto, stLearn, SpaGene, SpaGCN, BayesSpace, UTAG, STAGATE, DeepST, and SEDR). B Barplot for the 6 performance metrics (ARI, AMI, purity score, homogeneity, completeness, v measure) on the clustering results of the DLPFC 12 sections. Notation for statistical significance testing: n.s.p-value > 0.05, *p-value < 0.05, **p-value < 0.005, ***p-value < 0.0005, ****p-value < 0.00005. C UMAP and spatial visualization of Monocle 3 pseudo-time trajectories inferred with the latent representation by the tested methods of DLPFC slice #151673. UMAP plots with ground-truth labels (above), UMAP plots overlaid with Monocle 3 pseudo-time trajectories (middle), and Monocle 3 pseudo-time ordering on spatial coordinates (bottom). D The 12 human DLPFC sections with manual annotation. E Normalized cLISI and iLISI scores for DLPFC section integration results using the latent representations of four methods (Seurat, Harmony, STAGATE, SEDR). F UMAPs of integration results for four methods