Skip to main content
Fig. 2 | Genome Biology

Fig. 2

From: BASS: multi-scale and multi-sample analysis enables accurate cell type clustering and spatial domain detection in spatial transcriptomic studies

Fig. 2

Comparison of different methods for spatial domain detection and cell type clustering in simulations on a single tissue section. Boxplots of ARI show the accuracy of different methods for (A–D) spatial domain detection and (E–H) cell type clustering. Compared methods for spatial domain detection include BASS, HMRF, BayesSpace, and SpaGCN. For HMRF, a list of the spatial parameter βs ranging from 0 to 50 at increments of 2 were examined, and the three β values that corresponded to the worst, median, and best performance are displayed. Compared methods for cell type clustering include BASS, Seurat, SC3, and FICT. I–L Boxplots show the estimated cell type proportions in each spatial domain across simulation replicates, where πcr indicates the proportion of cell type c in the spatial domain r. The red dashed lines indicate the true proportions. Simulations were conducted under Scenario I (A, E, I), Scenario II (B, F, J), Scenario III (C, G, K), or Scenario IV (D, H, L), with the simulation parameters set as the baseline setting: the number of genes (nGenes) = 200, the proportion of genes that were differentially expressed in each cell type versus the others (de. prob) = 0.2, and the DE gene strength (de. facloc) = 1.1

Back to article page