Skip to contents

Proportion Test

Source: R/RunProportionTest.R
RunProportionTest.Rd

RunProportionTest performs differential abundance testing for cell proportions. The function acts as a dispatcher and routes to one of the method-specific implementations: permutation, milo, sccoda, or propeller. "propeller" uses the optional speckle backend for pairwise empirical Bayes moderated tests; see RunPropeller for transformations and parameters.

Usage

RunProportionTest(
  object,
  group.by,
  split.by = NULL,
  comparison = NULL,
  proportion_method,
  sample.by = NULL,
  pseudo_sample_n = 3L,
  allow_pseudo_samples = FALSE,
  n_permutations = 1000,
  FDR_threshold = 0.05,
  log2FD_threshold = log2(1.5),
  include_all_cells = FALSE,
  seed = 11,
  verbose = TRUE,
  ...,
  srt = NULL
)

Arguments

object

A Seurat object.

group.by

Metadata column(s) used to color cells.

split.by

Metadata column that identifies the condition groups to compare. For sample-level methods, if split.by is omitted and sample.by is provided, sample.by is treated as the condition column; a separate biological sample column is still required unless descriptive virtual samples are explicitly enabled.

comparison

Optional pairs of condition labels. Each pair is c(group1, group2). obs_log2FD is log2(group1 / group2), matching RunDEtest / Seurat ident.1 vs ident.2: positive values mean the cell type is more abundant in group1. If NULL, all pairwise comparisons are generated in both directions.

proportion_method

Differential abundance method. One of "permutation", "milo", "sccoda", or "propeller". This argument is required. Alias values such as "permutation_test" and "perm" are accepted and normalized to "permutation".

sample.by

Metadata column that identifies biological samples. Required for "milo", "sccoda", and "propeller" unless allow_pseudo_samples = TRUE. For Milo and Propeller, sample IDs may recur across two conditions for a fully paired design; partially paired comparisons are rejected.

pseudo_sample_n

Number of virtual samples per split.by group when a sample-level method has no usable sample.by and allow_pseudo_samples = TRUE. Virtual samples are descriptive only; inferential statistics are set to NA.

allow_pseudo_samples

Explicitly allow descriptive virtual samples when biological sample IDs are unavailable. The default is FALSE because partitioning cells does not create biological replicates.

n_permutations

Number of permutations for permutation-based test.

FDR_threshold

FDR value cutoff for significance.

log2FD_threshold

Absolute value of log2FD cutoff for significance.

include_all_cells

Whether to include all cell types in the complete grid for permutation mode.

seed

Random seed, including for permutation testing.

verbose

Whether to print the message. Default is TRUE.

...

Additional arguments passed to the selected method function.

srt

Deprecated alias for object; supply exactly one of the two. It will be removed in scop 1.0.0.

References

Miller SA, Policastro RA, Sriramkumar S, et al. (2021). LSD1 and Aberrant DNA Methylation Mediate Persistence of Enteroendocrine Progenitors That Support BRAF-Mutant Colorectal Cancer. Cancer Research, 81(14), 3791-3805. doi:10.1158/0008-5472.CAN-20-3562

scProportionTest, miloR, scCODA, propeller/speckle

See also

RunPermutation, RunMilo, RunscCODA, RunPropeller, ProportionTestPlot

Examples

data(pancreas_sub)
pancreas_sub <- RunProportionTest(
  pancreas_sub,
  group.by = "CellType",
  split.by = "Phase",
  proportion_method = "permutation",
  comparison = list(c("G2M", "G1"))
)
#> ℹ [2026-10-11 23:42:00] Start proportion test ("permutation")
#> ℹ [2026-10-11 23:42:00] Running comparison: "G2M" vs "G1"
#> ℹ [2026-10-11 23:42:00] Running comparison: "G1" vs "G2M"
#> ✔ [2026-10-11 23:42:00] Proportion test completed ("permutation")

ProportionTestPlot(
  pancreas_sub
)