SingleR是一种用于单细胞RNA测序(scRNAseq)数据的自动标注方法(Aran et al. 2019)。给定一个具有已知标签的参考样本集(单细胞或批量)，它根据与参考的相似性对来自测试数据集的新单元格进行标记。因此，对于参考数据集，手动解释集群和定义标记基因的负担只需要做一次，并且这种生物知识可以以自动的方式传播到新的数据集。singleR自带的7个参考数据集，其中5个是人类数据，2个是小鼠的数据：

1. BlueprintEncodeData Blueprint (Martens and Stunnenberg 2013) and Encode (The ENCODE Project Consortium 2012) （人）

2. DatabaseImmuneCellExpressionData The Database for Immune Cell Expression(/eQTLs/Epigenomics)(Schmiedel et al. 2018)（人）

3. HumanPrimaryCellAtlasData the Human Primary Cell Atlas (Mabbott et al. 2013)（人）

4. MonacoImmuneData, Monaco Immune Cell Data - GSE107011 (Monaco et al. 2019)（人）

5. NovershternHematopoieticData Novershtern Hematopoietic Cell Data - GSE24759（人）

ImmGenData the murine ImmGen (Heng et al. 2008) （鼠）

MouseRNAseqData a collection of mouse data sets downloaded from GEO (Benayoun et al. 2019).鼠）

为了简明扼要，本文只简要介绍了SingleR的基本功能。然而，该包还提供了更高级的功能，包括同时使用多个引用，操作细胞本体和提高大数据集的性能。读者可以参考这本书了解更多细节。

软件安装

if(!require(celldex))
  BiocManager::install("celldex")
if(!require(SingleR))
  BiocManager::install("SingleR")

数据读取

这里我们同样使用 SCS【4】单细胞转录组数据可视化分析 (Seurat 4.0) 数据例子，具体数据的处理过程这里都有，直接运行即可，此外关于降维聚类我们三种方法都有分析，三个不同是函数处理降维：RunPCA(), RunUMAP(), RunTSNE()。当我们输出pbmc时，可以看到下面一行信息，"3 dimensional reductions calculated: pca, umap, tsne"，说明三种降维方法都已完成。

library(Seurat)
# 进行singleR注释
pbmc.data <- Read10X(data.dir = "./filtered_gene_bc_matrices/hg19")
# Initialize the Seurat object with the raw (non-normalized data).
pbmc <- CreateSeuratObject(counts = pbmc.data, project = "pbmc3k", min.cells = 3,
    min.features = 200)
pbmc
## An object of class Seurat 
## 13714 features across 2700 samples within 1 assay 
## Active assay: RNA (13714 features, 0 variable features)
pbmc[["percent.mt"]] <- PercentageFeatureSet(pbmc, pattern = "^MT-")
pbmc <- NormalizeData(pbmc, normalization.method = "LogNormalize", scale.factor = 10000)
pbmc <- NormalizeData(pbmc)
pbmc <- FindVariableFeatures(pbmc, selection.method = "vst", nfeatures = 2000)
all.genes <- rownames(pbmc)
pbmc <- ScaleData(pbmc, features = all.genes)
pbmc <- RunPCA(pbmc, features = VariableFeatures(object = pbmc))
pbmc <- FindNeighbors(pbmc, dims = 1:10)
pbmc <- FindClusters(pbmc, resolution = 0.5)
## Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck
## 
## Number of nodes: 2700
## Number of edges: 97892
## 
## Running Louvain algorithm...
## Maximum modularity in 10 random starts: 0.8719
## Number of communities: 9
## Elapsed time: 0 seconds
pbmc <- RunUMAP(pbmc, dims = 1:10)
pbmc <- RunTSNE(pbmc, dims = 1:10)
pbmc
## An object of class Seurat 
## 13714 features across 2700 samples within 1 assay 
## Active assay: RNA (13714 features, 2000 variable features)
##  3 dimensional reductions calculated: pca, umap, tsne

绘制三种不同降维聚类的结果，如下：

plot1 <- DimPlot(pbmc, reduction = "pca", label = TRUE)
plot2 <- DimPlot(pbmc, reduction = "tsne", label = TRUE)
plot3 <- DimPlot(pbmc, reduction = "umap", label = TRUE)
plot1 + plot2 + plot3

数据库加载

我们选取两个 human相关的单细胞类型数据库，我们加载之后可以将变量保存起来，以免下次使用比较耗时，保存之后直接load即可，如下：

library(SingleR)
library(celldex)
### 数据库加载 hpca.se <- celldex::HumanPrimaryCellAtlasData() hpca.se@metadata
### bpe.se<-celldex::BlueprintEncodeData() bpe.se@metadata
### save(hpca.se,file='HumanPrimaryCellAtlas_hpca.se_human.RData')
### save(bpe.se,file='BlueprintEncode_bpe.se_human.RData')
load("HumanPrimaryCellAtlas_hpca.se_human.RData")
load("BlueprintEncode_bpe.se_human.RData")

单个数据库注释

只选择单个数据库，类似我们的细胞具有哪一类细胞的倾向性。

pbmc4SingleR <- GetAssayData(pbmc, slot = "data")  ##获取标准化矩阵
pbmc.hesc <- SingleR(test = pbmc4SingleR, ref = hpca.se, labels = hpca.se$label.main)  #
pbmc.hesc
## DataFrame with 2700 rows and 5 columns
##                                           scores     first.labels
##                                         <matrix>      <character>
## AAACATACAACCAC-1 0.1031269:0.234280:0.222822:...          T_cells
## AAACATTGAGCTAC-1 0.0984977:0.361135:0.300626:...           B_cell
## AAACATTGATCAGC-1 0.0671635:0.266281:0.237642:...          T_cells
## AAACCGTGCTTCCG-1 0.0836640:0.235213:0.273059:...         Monocyte
## AAACCGTGTATGCG-1 0.0739783:0.164291:0.175458:...          NK_cell
## ...                                          ...              ...
## TTTCGAACTCTCAT-1 0.0965528:0.249749:0.307289:... Pre-B_cell_CD34-
## TTTCTACTGAGGCA-1 0.1368350:0.302956:0.269191:... Pre-B_cell_CD34-
## TTTCTACTTCCTCG-1 0.0749830:0.272021:0.225157:...           B_cell
## TTTGCATGAGAGGC-1 0.0673739:0.232823:0.184013:...           B_cell
## TTTGCATGCCTCAC-1 0.0890275:0.242453:0.243141:...          T_cells
##                       tuning.scores           labels    pruned.labels
##                         <DataFrame>      <character>      <character>
## AAACATACAACCAC-1 0.301975:0.2133145          T_cells          T_cells
## AAACATTGAGCTAC-1 0.326636:0.0220241           B_cell           B_cell
## AAACATTGATCAGC-1 0.307280:0.1090948          T_cells          T_cells
## AAACCGTGCTTCCG-1 0.286386:0.2360274         Monocyte         Monocyte
## AAACCGTGTATGCG-1 0.299925:0.2107053          NK_cell          NK_cell
## ...                             ...              ...              ...
## TTTCGAACTCTCAT-1  0.259345:0.134353         Monocyte         Monocyte
## TTTCTACTGAGGCA-1  0.157336:0.129647 Pre-B_cell_CD34- Pre-B_cell_CD34-
## TTTCTACTTCCTCG-1  0.235417:0.173613           B_cell           B_cell
## TTTGCATGAGAGGC-1  0.228548:0.149524           B_cell           B_cell
## TTTGCATGCCTCAC-1  0.288418:0.194118          T_cells          T_cells

绘制降维聚类图

PCA 降维聚类

# seurat 和 singleR的table表
table(pbmc.hesc$labels, pbmc$seurat_clusters)
##                   
##                      0   1   2   3   4   5   6   7   8
##   B_cell             0   0   0 339   0   0   2   0   0
##   CMP                2   0   2   0   0   0   0   0   0
##   Monocyte           0 444   0   0   0 152   0  31   2
##   NK_cell            4   0   1   0  54   0 139   0   0
##   Platelets          0   0   0   0   0   0   0   0  12
##   Pre-B_cell_CD34-   4  47   1   1   1   6   2   4   0
##   Pro-B_cell_CD34+   0   0   0   4   0   0   0   0   1
##   T_cells          686   0 463   5 284   1   5   1   0
pbmc@meta.data$labels <- pbmc.hesc$labels
DimPlot(pbmc, group.by = c("seurat_clusters", "labels"), reduction = "pca")

t-SNE 降维聚类

DimPlot(pbmc, group.by = c("seurat_clusters", "labels"),reduction = "tsne")

UMAP 降维聚类

DimPlot(pbmc, group.by = c("seurat_clusters", "labels"),reduction = "umap")

多个数据库注释

使用BP和HPCA两个数据库综合注释，使用list函数读入多个数据库。

PCA 降维聚类

pbmc.hesc <- SingleR(test = pbmc4SingleR, ref = list(BP = bpe.se, HPCA = hpca.se),
    labels = list(bpe.se$label.main, hpca.se$label.main))
table(pbmc.hesc$labels, pbmc$seurat_clusters)
##                   
##                      0   1   2   3   4   5   6   7   8
##   B-cells            0   0   0 340   0   0   2   1   1
##   B_cell             0   0   0   4   0   0   1   0   0
##   CD4+ T-cells     298   0 159   1   1   0   0   0   0
##   CD8+ T-cells     376   0 282   3 294   0   7   1   0
##   HSC                2   0   2   0   0   0   0   1   0
##   Monocyte           0 297   0   0   0 125   0  25   1
##   Monocytes          0 181   0   0   0  32   0   8   1
##   NK cells           2   0   0   0  44   0 138   0   0
##   Platelets          0   0   0   0   0   0   0   0  12
##   Pre-B_cell_CD34-   0  13   0   0   0   1   0   0   0
##   T_cells           18   0  24   1   0   1   0   0   0

pbmc@meta.data$labels <- pbmc.hesc$labels

DimPlot(pbmc, group.by = c("seurat_clusters", "labels"), reduction = "pca")

t-SNE 降维聚类

DimPlot(pbmc, group.by = c("seurat_clusters", "labels"),reduction = "tsne")

UMAP 降维聚类

DimPlot(pbmc, group.by = c("seurat_clusters", "labels"),reduction = "umap")

注释结果诊断

1. 基于scores within cells

细胞在一个标签的得分很显著的高于其他标签得分，注释结果比较清晰。

### Annotation diagnostics
plotScoreHeatmap(pbmc.hesc)

2. 基于 per-cell “deltas”诊断

Delta值低，说明注释结果不是很明确。

plotDeltaDistribution(pbmc.hesc, ncol = 3)

3. 与cluster结果比较

tab <- table(label = pbmc.hesc$labels, cluster = pbmc$seurat_clusters)
tab
##                   cluster
## label                0   1   2   3   4   5   6   7   8
##   B-cells            0   0   0 340   0   0   2   1   1
##   B_cell             0   0   0   4   0   0   1   0   0
##   CD4+ T-cells     298   0 159   1   1   0   0   0   0
##   CD8+ T-cells     376   0 282   3 294   0   7   1   0
##   HSC                2   0   2   0   0   0   0   1   0
##   Monocyte           0 297   0   0   0 125   0  25   1
##   Monocytes          0 181   0   0   0  32   0   8   1
##   NK cells           2   0   0   0  44   0 138   0   0
##   Platelets          0   0   0   0   0   0   0   0  12
##   Pre-B_cell_CD34-   0  13   0   0   0   1   0   0   0
##   T_cells           18   0  24   1   0   1   0   0   0
library(pheatmap)
pheatmap(log10(tab + 10))

所有的可视化分析就基本完成了，学会了单细胞分析就非常简单了，目前测序的费用也在降低，单细胞系列可算是目前的测序神器，有这方面需求的老师，联系桓峰基因，提供最高端的科研服务！

桓峰基因，铸造成功的您！

未来桓峰基因公众号将不间断的推出单细胞系列生信分析教程，

敬请期待！！

有想进生信交流群的老师可以扫最后一个二维码加微信，备注“单位+姓名+目的”，有些想发广告的就免打扰吧，还得费力气把你踢出去！

References:

Aran, D., A. P. Looney, L. Liu, E. Wu, V. Fong, A. Hsu, S. Chak, et al. 2019. “Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage.” Nat. Immunol. 20 (2): 163–72.
Grun, D., M. J. Muraro, J. C. Boisset, K. Wiebrands, A. Lyubimova, G. Dharmadhikari, M. van den Born, et al. 2016. “De Novo Prediction of Stem Cell Identity using Single-Cell Transcriptome Data.” Cell Stem Cell 19 (2): 266–77.
La Manno, G., D. Gyllborg, S. Codeluppi, K. Nishimura, C. Salto, A. Zeisel, L. E. Borm, et al. 2016. “Molecular Diversity of Midbrain Development in Mouse, Human, and Stem Cells.” Cell 167 (2): 566–80.
Mabbott, Neil A., J. K. Baillie, Helen Brown, Tom C. Freeman, and David A. Hume. 2013. “An expression atlas of human primary cells: Inference of gene function from coexpression networks.” BMC Genomics 14. https://doi.org/10.1186/1471-2164-14-632.
Muraro, M. J., G. Dharmadhikari, D. Grun, N. Groen, T. Dielen, E. Jansen, L. van Gurp, et al. 2016. “A Single-Cell Transcriptome Atlas of the Human Pancreas.” Cell Syst 3 (4): 385–94.

继续滑动看下一个

SCS【6】单细胞转录组之细胞类型自动注释 (SingleR)

前 言

（1）人工注释

（2）软件自动化注释

软件安装

数据读取

数据库加载

单个数据库注释

PCA 降维聚类

t-SNE 降维聚类

UMAP 降维聚类

多个数据库注释

PCA 降维聚类

t-SNE 降维聚类

UMAP 降维聚类

注释结果诊断

1. 基于scores within cells

2. 基于 per-cell “deltas”诊断

3. 与cluster结果比较

References:

前言