Single-cell reference datasets for RCTD deconvolution in SpatialROI

This record contains single-cell RNA-seq references for cell-type deconvolution with RCTD in SpatialROI. Each Seurat object contains raw integer RNA counts and broad cell-type annotations in the cell_type metadata column. The collection includes references for lung adenocarcinoma and lung squamous cell carcinoma (Human Lung Cell Atlas; Sikkema et al., Nature Medicine, 2023), breast cancer (GSE176078), hepatocellular carcinoma (GSE149614), renal cell carcinoma (GSE152938), oral squamous cell carcinoma (GSM5017022–GSM5017032), and mouse brain (GSE115746). The colorectal cancer reference (GSE132465) and the colorectal cancer liver metastasis reference are also built into SpatialROI. Published annotations were grouped into broad cell types suitable for conventional Visium data. Ambiguous or low-quality cells were removed, and large cell populations were downsampled. SpatialROI excludes cell types with fewer than 25 cells and uses at most 2,000 cells per cell type when building the RCTD reference. The lung references were derived from normal tissue and do not contain malignant cells. The mouse brain reference was generated using SMART-seq rather than droplet-based UMI sequencing, so its results should be interpreted with caution. To use a reference, download the file and upload it under "Upload my own reference" in the Cell Type Deconvolution panel. All files are below the public server's 150 MB upload limit. Please cite the original studies when using these references.