Skip to contents

Test neighbourhoods for differential abundance

test_nhoods.Rd

Performs differential abundance testing on single-cell neighbourhoods using edgeR's quasi-likelihood negative binomial framework. The function fits a generalised linear model to neighbourhood cell counts, tests for differential abundance between conditions, and applies spatial FDR correction to account for overlapping neighbourhoods. This implementation follows the approach described in Dann et al., using graph-based neighbourhoods to identify regions of significant compositional changes in single-cell data.

Usage

test_nhoods(
  x,
  design,
  design_df,
  coef = NULL,
  norm_method = c("TMM", "RLE", "logMS"),
  min_mean = 0,
  robust = TRUE,
  fdr_weighting = c("k-distance", "graph-overlap", "none")
)

# S3 method for class 'miloR'
test_nhoods(
  x,
  design,
  design_df,
  coef = NULL,
  norm_method = c("TMM", "RLE", "logMS"),
  min_mean = 0,
  robust = TRUE,
  fdr_weighting = c("k-distance", "graph-overlap", "none")
)

Arguments

x

miloR object for which to run the differential abundance analysis.

design

Formula for the experimental design

design_df

data.frame. Contains the metadata to be used for the generation of the model matrix.

coef

Optional string/integer. For more complex experimental designs, you can specify which coefficient to test. If NULL, tests the last coefficient in the design matrix (typically the main effect of interest).

norm_method

String. Normalisation method to use. One of c("TMM", "RLE", "logMS"). Defaults to TMM (trimmed mean of M-values).

min_mean

Numeric. Minimum mean count threshold for filtering neighbourhoods. Neighbourhoods with mean counts below this value are excluded. Defaults to 0 (no filtering).

robust

Logical. If TRUE, uses robust estimation of the quasi-likelihood dispersion. Recommended for datasets with potential outliers. Defaults to TRUE.

fdr_weighting

String. Spatial FDR weighting scheme. One of c("k-distance", "graph-overlap", "none"). k-distance uses the distance to the k-th nearest neighbour, graph-overlap uses neighbourhood overlap counts. Defaults to k-distance.

Value

The miloR object with added model and results from the differential abundance analysis.

References

Dann et al., 2022, Nat Biotechnol