UCell: Robust and scalable single-cell gene signature scoring

UCell is an R package for scoring gene signatures in single-cell datasets. UCell scores, based on the Mann-Whitney U statistic, are robust to dataset size and heterogeneity, and their calculation demands relatively less computing time and memory than other robust methods, enabling the processing of large datasets (>10^5 cells). UCell can be applied to any cell vs. gene data matrix, and includes functions to directly interact with Seurat and Bioconductor's SingleCellExperiment objects.

Find the installation instructions for the package and usage vignettes below.

See also pyUCell for a Python implementation of UCell.

Package Installation

UCell is on Bioconductor To install the package from Bioc run:

if (!requireNamespace("BiocManager", quietly=TRUE))
    install.packages("BiocManager")
BiocManager::install("UCell")

For previous releases of UCell, you may download a tagged version from GitHub:

library(remotes)
remotes::install_github("carmonalab/UCell", ref="v2.2")

Test the package

Load sample data and test your installation:

library(UCell)

data(sample.matrix)
gene.sets <- list(Tcell_signature = c("CD2","CD3E","CD3D"),
			Myeloid_signature = c("SPI1","FCER1G","CSF1R"))

scores <- ScoreSignatures_UCell(sample.matrix, features=gene.sets)
head(scores)

Vignettes and examples

Vignettes to run UCell on matrices, SingleCellExperiment or Seurat objects can be found at the UCell Bioc page.

1. Gene signature scoring with UCell
2. Using UCell with SingleCellExperiment
3. Using UCell with Seurat
4. Some important UCell parameters

Additional tutorials are also available at:

• UCell vs. AddModuleScore()
• Using UCell and Seurat to identify different T cell subtypes in human tumors

New in version >= 2.7.6

Calculation of UCell scores has been updated as follows:

${UCell} = 1 - {U} / {U_{max}}$

where

$U = \sum_{i=1}^{n} r_i - \frac{n(n+1)}{2}$

and the normalization factor $U_{max}$ is:

$U_{max} = n \cdot {maxRank} - \frac{n(n+1)}{2}$

$n$ is the number of genes in the signature, $r_i$ are the cell-wise ranks for each of the $n$ genes, ${maxRank}$ is a parameter capping the ranking to the top genes (1500 by default), and $U$ is the Mann-Whitney U statistic (bounded by 0 and $U_{max}$). Earlier implementations used $U_{max} = n \cdot {maxRank}$ to normalize the U statistics. While for typical applications results should be similar between the two implementations, the new normalization provides more homogeneous UCell score distributions for large gene sets.

New in version >= 2.1.2

Single-cell data are sparse. It can be useful to 'impute' scores by neighboring cells and partially correct this sparsity. The new function SmoothKNN performs smoothing of single-cell signature scores by weighted average of the k-nearest neighbors in a given dimensionality reduction. It can be applied directly on SingleCellExperiment or Seurat objects to smooth UCell scores:

obj <- SmoothKNN(obj, signature.names = sigs, reduction="pca")

Interacting with signatures

For easy retrieval and storing of signatures, check out SignatuR:

remotes::install_github("carmonalab/SignatuR")
library(SignatuR)
#e.g. get a cycling signature
cycling.G1S <- GetSignature(SignatuR$Hs$Programs$cellCycle.G1S)

Note that UCell supports positive and negative gene sets within a signature. Simply append + or - signs to the genes to include them in positive and negative sets, respectively. For example:

my_signature <- c("CD2+","CD8A+","CD4-")

Get help

See more information about UCell and its functions by typing ?UCell within R. Please address your questions and bug reports at: UCell issues.

Citation

UCell: robust and scalable single-cell gene signature scoring. Massimo Andreatta & Santiago J Carmona (2021) CSBJ https://doi.org/10.1016/j.csbj.2021.06.043