Differential expression between cell clusters — FindClusterMarkers • Coralysis

FindClusterMarkers enables identifying feature markers for one cluster or two arbitrary combinations of clusters, e.g. 1_2 vs. 3_4_5. Feature and cell filters can be applied to accelerate the analysis, but this might lead to missing weak signals.

Usage

FindClusterMarkers.SingleCellExperiment(
  object,
  clustering.label,
  clusters.1,
  clusters.2,
  test,
  log2fc.threshold,
  min.pct,
  min.diff.pct,
  min.cells.group,
  max.cells.per.cluster,
  return.thresh,
  only.pos
)

# S4 method for class 'SingleCellExperiment'
FindClusterMarkers(
  object,
  clustering.label,
  clusters.1 = NULL,
  clusters.2 = NULL,
  test = "wilcox",
  log2fc.threshold = 0.25,
  min.pct = 0.1,
  min.diff.pct = NULL,
  min.cells.group = 3,
  max.cells.per.cluster = NULL,
  return.thresh = 0.01,
  only.pos = FALSE
)

Arguments

object: A SingleCellExperiment object.
clustering.label: A variable name (of class character) available in the cell metadata colData(object) with the clustering labels (character or factor) to use.
clusters.1: a character or numeric vector denoting which clusters to use in the first group (named group.1 in the results)
clusters.2: a character or numeric vector denoting which clusters to use in the second group (named group.2 in the results)
test: Which test to use. Only "wilcoxon" (the Wilcoxon rank-sum test, AKA Mann-Whitney U test) is supported at the moment.
log2fc.threshold: Filters out features that have log2 fold-change of the averaged feature expression values below this threshold. Default is 0.25.
min.pct: Filters out features that have dropout rate (fraction of cells expressing a feature) below this threshold in both comparison groups Default is 0.1.
min.diff.pct: Filters out features that do not have this minimum difference in the dropout rates (fraction of cells expressing a feature) between the two comparison groups. Default is NULL.
min.cells.group: The minimum number of cells in the two comparison groups to perform the DE analysis. If the number of cells is below the threshold, then the DE analysis is not performed. Default is 3.
max.cells.per.cluster: The maximun number of cells per cluster if downsampling is performed to speed up the DE analysis. Default is NULL, i.e. no downsampling.
return.thresh: If only.pos=TRUE, then return only features that have the adjusted p-value (adjusted by the Bonferroni method) below or equal to this threshold. Default is 0.01.
only.pos: Whether to return only features that have an adjusted p-value (adjusted by the Bonferroni method) below or equal to the threshold. Default is FALSE.

Value

a data frame of the results if positive results were found, else NULL

Examples

# Import package
suppressPackageStartupMessages(library("SingleCellExperiment"))

# Create toy SCE data
batches <- c("b1", "b2")
set.seed(239)
batch <- sample(x = batches, size = nrow(iris), replace = TRUE)
sce <- SingleCellExperiment(assays = list(logcounts = t(iris[,1:4])),  
                            colData = DataFrame("Species" = iris$Species, 
                                               "Batch" = batch))
colnames(sce) <- paste0("samp", 1:ncol(sce))

# Markers between versicolor vs virginica
dge <- FindClusterMarkers(sce, clustering.label = "Species", 
                          clusters.1 = "versicolor", 
                          clusters.2 = "virginica")
#> testing cluster group.1
#> 4 features left after min.pct filtering
#> 4 features left after min.diff.pct filtering
#> 3 features left after log2fc.threshold filtering
dge
#>                   p.value  adj.p.value log2FC pct.1 pct.2 diff.pct       marker
#> Sepal.Length 5.869006e-07 2.347603e-06 -0.652     1     1        0 Sepal.Length
#> Petal.Length 9.133545e-17 3.653418e-16 -1.292     1     1        0 Petal.Length
#> Petal.Width  9.698045e-17 3.879218e-16 -0.700     1     1        0  Petal.Width