【UCell】gene signature enrichment based on the Mann-Whitney U statistics

最新推荐文章于 2025-09-27 16:42:58 发布

原创最新推荐文章于 2025-09-27 16:42:58 发布 · 878 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#r语言

gene set/signature scoring 专栏收录该内容

3 篇文章

订阅专栏

UCell是一个工具，最初设计用于通过细胞类型签名打分进行细胞注释，也可以用于其他基因集的打分。它在Seurat对象的元数据列中存储每个细胞的得分。提供的示例包括简单快速的使用方法和完整的Seurat标准流程结合UCell进行细胞类型和亚群鉴定的流程。

ucell的设计本意是通过cell type signature打分进行细胞注释，但一样可以用于其它gene set的打分

首先做好gene signature list，与seurat流程相整合。对每个细胞计算每个条目的得分，每个条目的得分会存储在seurat object- metadata里的列。

demo1: 简单快速用法

注意：ucell直接在count上走流程，可视化的时候才normalize scale等.但其他的demo都是先走标准流程。

library(remotes)
remotes::install_github("carmonalab/UCell")


library(UCell)
set.seed(123)

###加载示例数据集
library(GEOquery)
cached.object <- "demodata.rds"

if (!file.exists(cached.object)) {
  geo_acc <- "GSE115978"
  options(timeout = max(1000, getOption("timeout")))
  
  gse <- getGEO(geo_acc)
  getGEOSuppFiles(geo_acc)
  
  exp.mat <- read.csv(sprintf("%s/GSE115978_counts.csv.gz", geo_acc), header = T,
                      row.names = 1, sep = ",")
  
  saveRDS(exp.mat, cached.object)
} else {
  exp.mat <- readRDS(cached.object)
}

###define gene signatures
signatures <- list(Immune = c("PTPRC"), 
                   Macrophage = c("CTSB", "C1QB", "LAPTM5", "TYROBP", "PSAP", "C1QA", "HLA-DRA", "CTSD", "NPC2", "FCER1G"), 
                   Tcell = c("CD3D", "CD3E", "CD3G", "CD2"), 
                   Bcell = c("MS4A1", "CD79A", "CD79B", "CD19", "BANK1"),
                   Myeloid_cell = c("CD14", "LYZ", "CSF1R", "FCER1G", "SPI1", "LCK-"), 
                   Stromal = c("MMP2", "COL1A1", "COL1A2", "COL3A1", "LUM", "DCN"))

###perform ucell
u.scores <- ScoreSignatures_UCell(exp.mat, features = signatures)
u.scores[1:8, 1:2]


###show the distrobution of the 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

###ucell seurat
library(Seurat)
seurat.object <- CreateSeuratObject(counts = exp.mat, project = "JerbyArnon")
seurat.object <- AddModuleScore_UCell(seurat.object, features = signatures, name = NULL,
                                      ncores = 2)
head(seurat.object@meta.data)


###generate pca and umap embeddings
seurat.object <- NormalizeData(seurat.object)
seurat.object <- FindVariableFeatures(seurat.object, selection.method = "vst", nfeatures = 500)

seurat.object <- ScaleData(seurat.object)
seurat.object <- RunPCA(seurat.object, features = seurat.object@assays$RNA@var.features,
                        npcs = 20)
seurat.object <- RunUMAP(seurat.object, reduction = "pca", dims = 1:20, seed.use = 123)

FeaturePlot(seurat.object, reduction = "umap", features = names(signatures), ncol = 3,
            order = T)

demo3: 完整地展示了seurat标准流程+ucell打分坚定细胞类型+亚群的流程

library(Seurat)
library(UCell)
library(dplyr)
set.seed(123)

inputFile <- "Yost.pretreatment.all.rds"
if (!file.exists(inputFile)) {
  download.file("https://drive.switch.ch/index.php/s/cluBLHkFFzLZWzL/download",
                inputFile)
}
data.seurat <- readRDS(inputFile)
data.seurat <- data.seurat %>%
  NormalizeData %>%
  FindVariableFeatures %>%
  ScaleData %>%
  RunPCA %>%
  RunUMAP(dims = 1:30)

data.seurat

### unsupervised clustering
data.seurat <- FindNeighbors(data.seurat, reduction = 'pca', dims = 1:30)
data.seurat <- FindClusters(data.seurat, resolution = 0.7)

DimPlot(data.seurat, reduction = "umap", group.by = "seurat_clusters", label = T) +
  NoLegend()

###cell signature
###from https://www.nature.com/articles/s41586-020-2157-4
signaturesHumanCellTypes <- list()

signaturesHumanCellTypes$Fetal.epithelial.progenitor <- c("BEX3" ,"STMN1",  "SOX4",  "LDHB" , "SKP1", "SNRPE" ,  "ID3" , "SRP9", "GSTP1" ,"SRP14" )

signaturesHumanCellTypes$Macrophage <- c("CTSB" ,   "C1QB",  "LAPTM5",  "TYROBP",    "PSAP"  ,  "C1QA", "HLA-DRA",    "CTSD" ,"NPC2" , "FCER1G" )

signaturesHumanCellTypes$B.cell..Plasmocyte. <- c("JCHAIN" ,  "IGHA1" ,   "SSR4"  ,  "MZB1" ,   "IGKC" ,  "IGHA2", "HERPUD1" ,  "DERL3" , "SEC11C" , "FKBP11" )

signaturesHumanCellTypes$Fibroblast <- c("C1S" , "TIMP2", "COL6A3" ,"SEMA3C" ,  "MMP2" ,   "GSN" ,"IGFBP6",  "MFAP4", "COL6A1" , "PLAC9")

signaturesHumanCellTypes$Fasciculata.cell <- c("PEBP1" ,   "STAR", "RARRES2" ,    "CLU" ,"CYP21A2", "CYP17A1" , "AKR1B1" ,    "NOV", "TPD52L1" ,  "EPHX1" )

signaturesHumanCellTypes$T.cell <- c("CD3D" , "CD3E" , "CD3G" ,  "CD4" ,  "CD2" ,  "CD7" , "TRAC", "TRBC1" ,  "LAT")

### ucell
data.seurat <- AddModuleScore_UCell(data.seurat, features = signaturesHumanCellTypes,
                                    ncores = 2)

# Some major cell types to look at:
toplot <- c("Fetal.epithelial.progenitor", "Macrophage", "B.cell..Plasmocyte.", "Fibroblast", "T.cell", "Fasciculata.cell"
            )

featnames <- paste0(toplot, "_UCell")
FeaturePlot(data.seurat, features = featnames, pt.size = 0.1, max.cutoff = "q99",
            ncol = 2)
VlnPlot(data.seurat, features = featnames, pt.size = 0, split.by = "seurat_clusters",
        ncol = 2)

###T.cell_UCell score > 0.2 as threshold
### identify t cells and subset them

# select as Tcell clusters only those with median Uscore>0.2
medians <- sapply(levels(data.seurat$seurat_clusters), function(x) {
  median(data.seurat@meta.data[data.seurat$seurat_clusters == x, "T.cell_UCell"])
})
tcell.clusters <- names(medians[medians > 0.2])

# Add metadata
data.seurat$is.Tcell <- FALSE
data.seurat@meta.data[data.seurat$seurat_clusters %in% tcell.clusters, "is.Tcell"] <- TRUE
DimPlot(data.seurat, group.by = "is.Tcell")

#subset on t cells
data.seurat.tcells <- subset(data.seurat, subset = is.Tcell == TRUE)
data.seurat.tcells

#recalculate embedding only for t cells
data.seurat.tcells <- data.seurat.tcells %>%
  NormalizeData %>%
  FindVariableFeatures %>%
  ScaleData %>%
  RunPCA %>%
  RunUMAP(dims = 1:20)

###Score tumor-infiltrating t cell subtype-specific signatures using UCell
#https://github.com/carmonalab/ProjecTILs
signaturesHumanTILs <- list()
signaturesHumanTILs$Tfh <- c("CD4" ,   "CD40LG" ,"TOX2" ,  "MAF" ,   "CD200" , "BATF"  ) 
signaturesHumanTILs$CD8_NaiveLike <- c("CD8A" , "CD8B" , "CCR7" , "IL7R" , "SELL" , "TCF7" , "S1PR1", "LEF1" )
signaturesHumanTILs$CD8_EffectorMemory <- c("CD8A",  "CD8B" , "GZMA" , "GZMK",  "CCL5" , "CXCR3")
signaturesHumanTILs$Thelper <- c("CD40LG", "CD4",    "IL7R"  , "RORA" ,  "ANXA1" )
signaturesHumanTILs$CD4_NaiveLike <- c("CD40LG" ,"CD4"  ,  "CCR7" ,  "SELL" ,  "IL7R" ,  "TCF7" ,  "LEF1"  )
signaturesHumanTILs$CD8_Tpex <- c("CD8A" , "CD8B" , "LAG3" , "XCL1" , "CRTAM" ,"TOX"  , "ZEB2" , "PDCD1" ,"TCF7", "CCR7" )
signaturesHumanTILs$CD8_Tex <- c("CD8A" ,  "CD8B" ,  "LAG3" ,  "HAVCR2", "GZMB" ,  "PRF1" ,  "PDCD1" , "TIGIT" )
signaturesHumanTILs$Treg<- c("CD4" ,  "IL2RA", "FOXP3")
signaturesHumanTILs[["cycling"]] <- c("TOP2A", "MKI67", "STMN1")


data.seurat.tcells <- AddModuleScore_UCell(data.seurat.tcells, features = signaturesHumanTILs,
                                           ncores = 2)

featnames <- paste0(names(signaturesHumanTILs), "_UCell")
FeaturePlot(data.seurat.tcells, features = featnames, pt.size = 0.1, order = T)
VlnPlot(data.seurat.tcells, features = featnames, pt.size = 0, split.by = "seurat_clusters")