differential_analysis_scExp: Runs differential analysis between cell clusters

In vallotlab/ChromSCape: Analysis of single-cell epigenomics datasets with a Shiny App

Runs differential analysis between cell clusters

Description

Based on clusters of cell defined previously, runs non-parametric Wilcoxon Rank Sum test to find significantly depleted or enriched features, in 'one_vs_rest' mode or 'pairwise' mode. In pairwise mode, each cluster is compared to all other cluster individually, and then pairwise comparisons between clusters are combined to find overall differential features using combineMarkers function from scran.

Usage

differential_analysis_scExp(
  scExp,
  de_type = c("one_vs_rest_fast", "one_vs_rest", "pairwise", "custom")[1],
  by = "cell_cluster",
  method = "wilcox",
  block = NULL,
  group = NULL,
  ref = NULL,
  prioritize_genes = nrow(scExp) > 20000,
  max_distanceToTSS = 1000,
  progress = NULL,
  BPPARAM = BiocParallel::bpparam()
)

Arguments

scExp	A SingleCellExperiment object containing consclust with selected number of cluster.
de_type	Type of comparisons. Either 'one_vs_rest', to compare each cluster against all others, or 'pairwise' to make 1 to 1 comparisons. ('one_vs_rest')
by	= A character specifying the column of the object containing the groups of cells to compare. Exclusive with de_type == custom
method	Differential testing method, either 'wilcox' for Wilcoxon non- parametric testing or 'neg.binomial' for edgerGLM based testing. ('wilcox')
block	Use batches as blocking factors ? If TRUE, block will be taken as the column "batch_id" from the SCE. Cells will be compared only within samples belonging to the same batch.
group	If de_type = "custom", the sample / cluster of interest as a one- column data.frame. The name of the column is the group name and the values are character either cluster ("C1", "C2", ...) or sample_id.
ref	If de_type = "custom", the sample / cluster of reference as a one- column data.frame. The name of the column is the group name and the values are character either cluster ("C1", "C2", ...) or sample_id.
prioritize_genes	First filter by loci being close to genes ? E.g. for differential analysis, it is more relevant to keep features close to genes
max_distanceToTSS	If prioritize_genes is TRUE, the maximum distance to consider a feature close to a gene.
progress	A shiny Progress instance to display progress bar.
BPPARAM	BPPARAM object for multiprocessing. See bpparam for more informations. Will take the default BPPARAM set in your R session.

Details

This functions takes as input a SingleCellExperiment object with consclust, the type of comparison, either 'one_vs_rest' or 'pairwise', the adjusted p-value threshold (qval.th) and the fold-change threshold (logFC.th). It outputs a SingleCellExperiment object containing a differential list.

Value

Returns a SingleCellExperiment object containing a differential list.

Examples

data("scExp")
scExp_cf = differential_analysis_scExp(scExp)

vallotlab/ChromSCape documentation built on Oct. 15, 2023, 1:47 p.m.