Fig. 2: Alignment and integrative analysis of three spatial datasets from breast cancer.

a, Registration of two breast cancer Xenium replicate (rep.) slides, one Visium slide and their corresponding H&E images to a CCS based on interactively selected landmarks. b, Illustration of how spatial annotations can be transferred across datasets using the CCS. From top to bottom, spatial annotations derived from multiple datasets, including histological regions (H&E image), tumor clones (Visium-derived copy number aberrations) and cell types (Xenium and scRNA-seq). Spatial annotations, represented by different spatial elements (polygons, circles, molecules), can can be transferred between datasets via the CCS. c, SpatialData queries facilitate cross-modality aggregation, quality control and benchmarking. Left and middle, cell-type fractions in Xenium computed at circular regions corresponding to Visium quantification locations; right, cell-type fraction estimates from deconvolution methods based on Visium data (using cell2location). d, Use of SpatialData queries for arbitrary geometrical quantifications. Shown are cell-type fraction estimates obtained in Xenium (derived from the paired scRNA-seq dataset) and Visium (cell2location estimates) at annotated ROIs and clones as in b. e, Comparison of gene expression quantification in Xenium and Visium using SpatialData aggregations at Visium capture locations. Left, scatter plot of the correlation coefficient of aggregated gene expression quantifications between Xenium replicates (x axis) versus that between Xenium and Visium (y axis). Shown are gene expression quantifications for 313 genes (dots) present in both Xenium and Visium. Color denotes log expression in Xenium replicate 1. Right, visualization of aggregated expression levels at Visium locations for FOXA1 (top) and UCP1 (bottom). Color bars denote raw counts.