SpotSweeper

Workflows

Standard Workflow

Identify and visualize spatially-aware local outliers based on library size, unique genes, and mitochondrial percentage.

library(SpotSweeper)
library(SpatialExperiment)

# 1. Load example data and drop out-of-tissue spots
spe <- STexampleData::Visium_humanDLPFC()
spe <- spe[, spe$in_tissue == 1]

# 2. Calculate QC metrics using scuttle
rownames(spe) <- rowData(spe)$gene_name
is.mito <- rownames(spe)[grepl("^MT-", rownames(spe))]
spe <- scuttle::addPerCellQCMetrics(spe, subsets = list(Mito = is.mito))

# 3. Identify local outliers
spe <- localOutliers(spe, metric = "sum", direction = "lower", log = TRUE)
spe <- localOutliers(spe, metric = "detected", direction = "lower", log = TRUE)
spe <- localOutliers(spe, metric = "subsets_Mito_percent", direction = "higher", log = FALSE)

# 4. Combine all outliers into "local_outliers" column
spe$local_outliers <- as.logical(spe$sum_outliers) | 
                      as.logical(spe$detected_outliers) | 
                      as.logical(spe$subsets_Mito_percent_outliers)

# 5. Visualize local outliers
library(escheR)
plotQCmetrics(spe, metric = "sum_log", outliers = "local_outliers", point_size = 1.1, stroke = 0.75) +
  ggtitle("All Local Outliers")

Input: A SpatialExperiment object with raw counts. Output: A SpatialExperiment object with identified local outliers annotated in colData.

Technical Artifact Detection

Identify and visualize technical artifacts (such as tissue hangnails) using local variance of mitochondrial metrics across multiple neighborhood sizes.

library(SpotSweeper)

# 1. Load data with artifact
data(DLPFC_artifact)
spe <- DLPFC_artifact

# 2. Visualize raw mitochondrial percentage
plotQCmetrics(spe, metric = "expr_chrM_ratio", outliers = NULL, point_size = 1.1) +
  ggtitle("Mitochondrial Percent")

# 3. Run findArtifacts to identify artifacts
spe <- findArtifacts(
  spe, 
  mito_percent = "expr_chrM_ratio", 
  mito_sum = "expr_chrM", 
  n_order = 5, 
  name = "artifact"
)

# 4. Visualize identified artifacts
plotQCmetrics(spe, metric = "expr_chrM_ratio", outliers = "artifact", point_size = 1.1) +
  ggtitle("Hangnail artifact")

Input: A SpatialExperiment object containing technical artifacts (e.g., tissue hangnails). Output: A SpatialExperiment object with artifact spots labeled in colData.

When to Use

• Detecting spot-level local outliers in spatial transcriptomics data based on library size, unique genes, and mitochondrial percentage using localOutliers().
• Identifying large technical artifacts (such as tissue hangnails) using local variance of mitochondrial metrics across multiple neighborhood sizes with findArtifacts().
• Visualizing spatial QC metrics and highlighting outliers or artifacts using plotQCmetrics().

When NOT to Use

• For non-spatial single-cell RNA-seq data, as the outlier detection methods rely on spatial coordinates and nearest-neighbor structures.
• When spatial coordinate metadata is missing from the input SpatialExperiment object.

Data Requirements

• A SpatialExperiment object containing spatial coordinates and count data.
• Pre-calculated QC metrics (such as sum, detected genes, and mitochondrial percentage) in the colData of the object (e.g., added via scuttle::addPerCellQCMetrics()).

Key Parameters

• metric: The QC metric column name in colData evaluated by localOutliers().
• direction: The direction of outlier detection ("lower" or "higher") in localOutliers().
• log: Logical flag indicating whether to log-transform the metric in localOutliers().
• mito_percent: Column name for mitochondrial percentage in findArtifacts().
• mito_sum: Column name for mitochondrial sum in findArtifacts().
• n_order: Neighborhood order size used to calculate local variance in findArtifacts().
• name: Column name to store the identified artifact logical vector in colData.
• outliers: Column name of outliers to highlight in plotQCmetrics().

Best Practices

• Filter out out-of-tissue spots (e.g., spe$in_tissue == 1) before running outlier detection.
• Log-transform highly skewed metrics like library size (sum) and unique genes (detected) by setting log = TRUE in localOutliers().
• Combine individual outlier logical vectors (e.g., sum_outliers, detected_outliers) into a single logical column for comprehensive visualization.

Common Pitfalls

• Including out-of-tissue spots: Running localOutliers() on background spots can skew the local neighborhood statistics; subset the object to in_tissue == 1 first.
• Using raw values for skewed metrics: Failing to log-transform library size or detected genes can lead to poor outlier detection; ensure log = TRUE is set for these metrics.

Alternatives

• scuttle: For standard non-spatial single-cell quality control metrics.
• escheR: For custom spatial visualization of transcriptomics data.

Citations

• Tott, M. (2026). Getting Started with SpotSweeper. R Package Vignette.

References

• Homepage: bioconductor.org/packages/spotsweeper
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/spotsweeper/inst/doc/getting_started.html