Fig. 3

Crescendo batch corrects technology effects between a colorectal cancer (CRC) scRNA-seq dataset and two CRC spatial transcriptomics samples. A Colorectal cancer samples were assayed with scRNA-seq and spatial transcriptomics. These datasets shared 477 genes. B UMAP embedding of cells from scRNA-seq and spatial transcriptomics before and after batch correction with Harmony (correction performed on a batch variable where the scRNA-seq dataset and each spatial slice was considered a batch). C Broad cell type classification of cells and spatial locations of cell types in spatial slices (middle, right). Gene expression distributions across slices for MS4A1 (D) and CD3D (G). E, H In these and following plots, scRNA-seq is plotted in UMAP space, while spatial slices are plotted in physical space. Spatial locations of cell types with the highest expression of MS4A1 (E) and CD3D (H). Gene expression visualizations in physical space before and after Crescendo batch correction for MS4A1 (F) and CD3D (I). J Scatter plots of batch-variance ratio (BVR) and cell-type-variance ratio (CVR) metrics calculated for all 477 genes across 5 different batch correction algorithms. Purple dashed vertical line is at CVR = 0.5 and the purple dashed horizontal line is at BVR = 1. Red at BVR < 1 and CVR ≥ 0.5 is the target zone for genes that were batch corrected well