1 Introduction

In single-cell RNA-seq analysis, gene signature (or “module”) scoring constitutes a simple yet powerful approach to evaluate the strength of biological signals, typically associated to a specific cell type or biological process, in a transcriptome.

UCell is an R package for evaluating gene signatures in single-cell datasets. UCell signature scores, based on the Mann-Whitney U statistic, are robust to dataset size and heterogeneity, and their calculation demands less computing time and memory than other available methods, enabling the processing of large datasets in a few minutes even on machines with limited computing power. UCell can be applied to any single-cell data matrix, and includes functions to directly interact with Seurat objects.

2 Quick start

To test your installation, load a small sample dataset and run UCell:

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)
##                       Tcell_signature_UCell Myeloid_signature_UCell
## L5_ATTTCTGAGGTCGTGA               0.8993333                       0
## L4_TCACTATTCATCTCTA               0.6904444                       0
## L1_TCCTTCTTCTTTACAC               0.5935556                       0
## L5_AAAGTGAAGGCGCTCT               0.6093333                       0
## E2L3_CCTCAGTAGTGCAGGT             0.8508889                       0
## L5_CCCTCTCGTTCTAAGC               0.6562222                       0

3 Get some testing data

For this demo, we will download a single-cell dataset of lung cancer (Zilionis et al. (2019) Immunity) through the scRNA-seq package. This dataset contains >170,000 single cells; for the sake of simplicity, in this demo will we focus on immune cells, according to the annotations by the authors, and downsample to 5000 cells.

library(scRNAseq)

lung <- ZilionisLungData()
immune <- lung$Used & lung$used_in_NSCLC_immune
lung <- lung[,immune]
lung <- lung[,1:5000]

exp.mat <- Matrix::Matrix(counts(lung),sparse = TRUE)
colnames(exp.mat) <- paste0(colnames(exp.mat), seq(1,ncol(exp.mat)))

4 Define gene signatures

Here we define some simple gene sets based on the “Human Cell Landscape” signatures Han et al. (2020) Nature. You may edit existing signatures, or add new one as elements in a list.

signatures <- list(
    Tcell = c("CD3D","CD3E","CD3G","CD2","TRAC"),
    Myeloid = c("CD14","LYZ","CSF1R","FCER1G","SPI1","LCK-"),
    NK = c("KLRD1","NCR1","NKG7","CD3D-","CD3E-"),
    Plasma_cell = c("MZB1","DERL3","CD19-")
)

5 Run UCell

Run ScoreSignatures_UCell and get directly signature scores for all cells

u.scores <- ScoreSignatures_UCell(exp.mat,features=signatures)
head(u.scores)
##         Tcell_UCell Myeloid_UCell NK_UCell Plasma_cell_UCell
## bcHTNA1           0     0.5234000        0             0.000
## bcHNVA2           0     0.5120000        0             0.000
## bcALZN3           0     0.3593333        0             0.076
## bcFWBP4           0     0.1558000        0             0.000
## bcBJYE5           0     0.4639333        0             0.000
## bcGSBJ6           0     0.5460000        0             0.000

Show the distribution of predicted scores

library(reshape2)
library(ggplot2)
melted <- reshape2::melt(u.scores)
colnames(melted) <- c("Cell","Signature","UCell_score")
p <- ggplot(melted, aes(x=Signature, y=UCell_score)) + 
    geom_violin(aes(fill=Signature), scale = "width") +
    geom_boxplot(width=0.1, outlier.size=0) +
    theme_bw() + theme(axis.text.x=element_blank())
p

6 Pre-calculating gene rankings

The time- and memory-demanding step in UCell is the calculation of gene rankings for each individual cell. If we plan to experiment with signatures, editing them or adding new cell subtypes, it is possible to pre-calculate the gene rankings once and for all and then apply new signatures over these pre-calculated ranks. Run the StoreRankings_UCell function to pre-calculate gene rankings over a dataset:

set.seed(123)
ranks <- StoreRankings_UCell(exp.mat)
ranks[1:5,1:5]
## 5 x 5 sparse Matrix of class "dgCMatrix"
##           bcHTNA1 bcHNVA2 bcALZN3 bcFWBP4 bcBJYE5
## 5S_rRNA         .       .       .       .       .
## 5_8S_rRNA       .       .       .       .       .
## 7SK             .       .       .       .       .
## A1BG            .       .       .       .       .
## A1BG-AS1        .       .       .       .       .

Then, we can apply our signature set, or any other new signature to the pre-calculated ranks. The calculations will be considerably faster.

set.seed(123)
u.scores.2 <- ScoreSignatures_UCell(features=signatures,
                                    precalc.ranks = ranks)

melted <- reshape2::melt(u.scores.2)
colnames(melted) <- c("Cell","Signature","UCell_score")
p <- ggplot(melted, aes(x=Signature, y=UCell_score)) + 
    geom_violin(aes(fill=Signature), scale = "width") +
    geom_boxplot(width=0.1, outlier.size = 0) + 
    theme_bw() + theme(axis.text.x=element_blank())
p

new.signatures <- list(Mast.cell = c("TPSAB1","TPSB2","CPA3","MS4A2"),
                       Lymphoid = c("LCK"))

u.scores.3 <- ScoreSignatures_UCell(features=new.signatures,
                                    precalc.ranks = ranks)
melted <- reshape2::melt(u.scores.3)
colnames(melted) <- c("Cell","Signature","UCell_score")
p <- ggplot(melted, aes(x=Signature, y=UCell_score)) + 
    geom_violin(aes(fill=Signature), scale = "width") +
    geom_boxplot(width=0.1, outlier.size=0) + 
    theme_bw() + theme(axis.text.x=element_blank())
p