Fig. 4: SpotClean improves marker specificity in human breast cancer sample human_breast_2.

a H&E image (left) and spots annotated as tumor vs. non-tumor via a pathologist’s visual inspection (right). Scale bar is 1 mm. b Expression of two tumor-specific markers in the raw (left) and SpotClean decontaminated (right) data. SpotClean increases specificity of these markers by maintaining expression in the tumor regions, and reducing expression in the non-tumor regions. c Boxplots of the marker expression shown in b as well as the marker expression in a breast cancer single-cell RNA-seq reference dataset²³. For the six boxplots from top to bottom in both ERBB2 and MUC1, n = 1843, 675, 1843, 675, 198, 317 spots or cells. The lower whisker, lower hinge, line inside box, upper hinge, and upper whisker represent the minimum, lower quartile, median, upper quartile, and maximum calculated without outlier values which are more than 1.5*inter-quartile range away from the hinges and are shown in separate dots. d UMAP plots generated from raw and SpotClean decontaminated data colored by spot annotations. SpotClean decontaminated data leads to improved separation of the groups, as shown visually, and quantified by the adjusted rand index (ARI) scores which show a 13% improvement in the SpotClean decontaminated data.