Cluster-aware spatial quality control for single-cell resolution spatial transcriptomics.

CellSweeper provides a QC framework for single-cell resolution spatial transcriptomics platforms such as Xenium, CosMx, MERFISH, VisiumHD, and others. It extends the SpotSweeper local outlier framework in control for confounding biology (Totty, Hicks & Guo, Nature Methods 2025), while also incorporating segmentation morphology metrics from SpaceTrooper.

The Three-Level QC Framework

Level	Function(s)	What it does
1. Global pre-filtering	globalFilter()	Remove obvious low quality observations — near-zero counts, impossible segmentations, etc
2. Cluster-level QC	clusterCellTypes() + flagArtifactClusters()	Cluster cells in gene expression space, then flag entire artifact clusters via multivariate pseudobulk outlier detection and spatial dispersion analysis
3. Within-cluster local outlier detection	clusterLocalOutliers()	Local outlier detection restricted to same-cluster neighbors, applied to both transcriptomic and morphological metrics

Installation

## Install from Bioconductor (coming soon)
# if (!requireNamespace("BiocManager", quietly = TRUE))
#     install.packages("BiocManager")
# BiocManager::install("CellSweeper")

# Or install the development version from GitHub
devtools::install_github("MicTott/CellSweeper")

Integration with SpaceTrooper

For imaging-based platforms (Xenium, CosMx, MERFISH), we recommend preprocessing with SpaceTrooper to compute morphology-based QC metrics and cell boundary polygons before running CellSweeper:

library(SpaceTrooper)
spe <- readXeniumSPE("/path/to/xenium/output/")
spe <- spatialPerCellQC(spe)    # adds Area_um, log2AspectRatio, etc.
spe <- addPolygonsToSPE(spe)    # adds sf polygons for visualization

library(CellSweeper)
spe <- runCellSweeper(spe,
    morpho_metrics = c("Area_um", "log2AspectRatio", "log2CountArea"))

Citation

If you use CellSweeper in your research, please cite:

Totty M, Hicks SC, Guo B. CellSweeper: spatially-aware quality control for single cell spatial transcriptomics. In preparation.

If you use SpaceTrooper for preprocessing, please also cite:

Add SpaceTrooper citation here:

CellSweeper is a direct extension of SpotSweeper:

Totty M, Hicks SC, Guo B. SpotSweeper: spatially aware quality control for spatial transcriptomics. Nature Methods 22, 1520–1530 (2025). doi:10.1038/s41592-025-02713-3

Report Bugs and Suggest Features.

Please open an issue to report bugs or suggest features.