Fig. 3: CellSPA graphical representation of comparison study using Xenium-BreastCancer. | Nature Communications

Fig. 3: CellSPA graphical representation of comparison study using Xenium-BreastCancer.

From: BIDCell: Biologically-informed self-supervised learning for segmentation of subcellular spatial transcriptomics data

Fig. 3: CellSPA graphical representation of comparison study using Xenium-BreastCancer.

a Correlation heatmap of average expression between segmented cells from BIDCell (y-axis) and expression from Chromium data (x-axis) [left]. Scatter plot between correlation with Chromium expression (y-axis) and average total number of transcripts per cell (x-axis) based on average expression [right]. Each dot represents a different method. b Scatter plot between correlation with Chromium expression (y-axis) and average total number of transcripts per cell (x-axis), where each dot represents a different method. c Scatter plot between BIDCell (y-axis) and expression from Chromium data (x-axis) based on the cell type proportion extracted from each of the methods. d Scatter plot showing the expression between the F1 score for positive markers in BIDCell (y-axis) and in 10x segmentation (x-axis) [left], and scatter plot showing the purity F1 score against the average total transcripts per cell [right]. Each dot represents a method. e Line plots showing the percentage of B cells expressing the unwanted T cell marker CD4, CD8A, and CD8B against its distance from the nearest T cell, where the B cells are grouped by distance ranges. A lower percentage is better, and each line represents a different method. f–h Spatial characteristics diversity. f indicates the local spatial regions being divided in the images where the left panel indicates the cell type proportions of each local region and the right panel indicates the cell type entropy of the local region. g Scatter plots showing the association between the cell type entropy and the coefficient of variation of the total transcripts of three methods: 10x, BIDCell, and Watershed, where each dot represents each local region shown in (f). h Scatter plots showing the association between the coefficient of variation of elongation and proportion of fibroblasts in the data. i Spatial imaging of two replicates in Xenium-BreastCancer, where each dot represents the segmented cells coloured by the annotated cell type. j UMAP plots of the two replicates, coloured by cell type [left] and replicate [right]. Source data are provided as a Source Data file.

Back to article page