AddModuleScore_UCell: Calculate module enrichment scores from single-cell data...

In carmonalab/UCell: Rank-based signature enrichment analysis for single-cell data

Calculate module enrichment scores from single-cell data (Seurat interface)

Description

Given a Seurat object, calculates module/signature enrichment scores at single-cell level using the Mann-Whitney U statistic. UCell scores are normalized U statistics (between 0 and 1), and they are mathematically related to the Area under the ROC curve (see Mason and Graham)

Usage

AddModuleScore_UCell(
  obj,
  features,
  maxRank = 1500,
  chunk.size = 100,
  BPPARAM = NULL,
  ncores = 1,
  storeRanks = FALSE,
  w_neg = 1,
  assay = NULL,
  slot = "counts",
  ties.method = "average",
  force.gc = FALSE,
  name = "_UCell"
)

Arguments

obj	Seurat object
features	A list of signatures, for example: list( Tcell_signature = c("CD2","CD3E","CD3D"), Myeloid_signature = c("SPI1","FCER1G","CSF1R")) You can also specify positive and negative gene sets by adding a + or - sign to genes in the signature; see an example below
maxRank	Maximum number of genes to rank per cell; above this rank, a given gene is considered as not expressed.
chunk.size	Number of cells to be processed simultaneously (lower size requires slightly more computation but reduces memory demands)
BPPARAM	A BiocParallel::bpparam() object that tells UCell how to parallelize. If provided, it overrides the ncores parameter.
ncores	Number of processors to parallelize computation. If BPPARAM = NULL, the function uses BiocParallel::MulticoreParam(workers=ncores)
storeRanks	Store ranks matrix in Seurat object ('UCellRanks' assay) for fast subsequent computations. This option may demand large amounts of RAM.
w_neg	Weight on negative genes in signature. e.g. w_neg=1 weighs equally up- and down-regulated genes, w_neg=0.5 gives 50% less importance to negative genes
assay	Pull out data from this assay of the Seurat object (if NULL, use DefaultAssay(obj))
slot	Pull out data from this slot (layer in v5) of the Seurat object
ties.method	How ranking ties should be resolved - passed on to data.table::frank
force.gc	Explicitly call garbage collector to reduce memory footprint
name	Name tag that will be appended at the end of each signature name, "_UCell" by default (e.g. signature score in meta data will be named: Myeloid_signature_UCell)

Details

In contrast to Seurat's AddModuleScore, which is normalized by binning genes of similar expression at the population level, UCell scores depend only on the gene expression ranks of individual cell, and therefore they are robust across datasets regardless of dataset composition.

Value

Returns a Seurat object with module/signature enrichment scores added to object meta data; each score is stored as the corresponding signature name provided in features followed by the tag given in name (or "_UCell" by default )

Examples

library(UCell)
gene.sets <- list(Tcell = c("CD2","CD3E","CD3D"),
                Myeloid = c("SPI1","FCER1G","CSF1R"))
data(sample.matrix)
obj <- Seurat::CreateSeuratObject(sample.matrix)                

obj <- AddModuleScore_UCell(obj,features = gene.sets)
head(obj[[]])

## Using positive and negative gene sets
gene.sets <- list()
gene.sets$Tcell_gd <- c("TRDC+","TRGC1+","TRGC2+","TRDV1+",
    "TRAC-","TRBC1-","TRBC2-")
gene.sets$NKcell <- c("FGFBP2+", "SPON2+", "KLRF1+",
    "FCGR3A+", "CD3E-","CD3G-")
obj <- AddModuleScore_UCell(obj, features = gene.sets, name=NULL)
head(obj$NKcell)

carmonalab/UCell documentation built on April 26, 2024, 11:51 p.m.