跟着大神学单细胞数据分析
前言
这是 Tang Ming 大神分享的单细胞分析的seurat流程。今天我们来理一下大致的分析思路，当然里面好多细节的部分还需要自己下功夫慢慢研究。
原文链接如下：
https://crazyhottommy.github.io/scRNA-seq-workshop-Fall-2019/scRNAseq_workshop_1.html
下载数据
我们将下载来自10x Genomics的公共 5k pbmc (外周血单核细胞)数据集。然后用R分析
1wget http://cf.10xgenomics.com/samples/cell-exp/3.0.2/5k_pbmc_v3/5k_pbmc_v3_filtered_feature_bc_matrix.tar.gz
2
3tar xvzf 5k_pbmc_v3_filtered_feature_bc_matrix.tar.gz
安装所需的R包
1install.packages("tidyverse")
2install.packages("rmarkdown")
3install.packages('Seurat')
如果你已经安装过这写R包，你可以忽略这一步。如果还没有安装或者安装R包有问题，可以参考下面的教程：
rstudio软件无需联网但是
BiocManger无法安装R包
批量安装R包小技巧大放送
读入数据
1# 读取PBMC数据集
2pbmc.data <- Read10X(data.dir = "filtered_feature_bc_matrix/")
3# 使用原始数据（未归一化处理）初始化Seurat对象
4pbmc <- CreateSeuratObject(counts = pbmc.data, project = "pbmc5k", min.cells = 3, min.features = 200)
5pbmc
61An object of class Seurat
218791 features across 4962 samples within 1 assay
3Active assay: RNA (18791 features)
如果你想了解更多Seurat对象的详细信息，你可以参考这个网站：https://github.com/satijalab/seurat/wiki
注：读入数据这一步使用的Seurat包应该是 Seurat V3版本。因为我用Seurat V2创建的对象和文中所给的结果不一致
1## 使用Srurat V2 创建对象
2pbmc <- CreateSeuratObject(raw.data = pbmc.data, project = "pbmc5k", min.cells = 3, min.features = 200)
3
4pbmc
5
6An object of class seurat in project pbmc5k
7 18791 genes across 5025 samples.
质量控制
1## check at metadata
2head(pbmc@meta.data)
3# The [[ operator can add columns to object metadata. This is a great place to stash QC stats
4pbmc[["percent.mt"]] <- PercentageFeatureSet(pbmc, pattern = "^MT-")
5pbmc@meta.data %>% head()
6
7##将质量控制指标可视化为小提琴图
8VlnPlot(pbmc, features = c("nFeature_RNA", "nCount_RNA", "percent.mt"), ncol = 3)
9
10#我们根据上面的可视化设置了截止值。这个截止值是相当主观的。
11pbmc <- subset(pbmc, subset = nFeature_RNA > 200 & nFeature_RNA < 5000 & percent.mt < 25)
Normalization
通常情况下，我们采用全局缩放的归一化方法"LogNormalize"
1pbmc <- NormalizeData(pbmc, normalization.method = "LogNormalize", scale.factor = 10000)
2
不过，现在Seurat也有一个新的标准化的方法，称为SCTransform . 详细了解可以查看：https://satijalab.org/seurat/v3.0/sctransform_vignette.html
特征选择
1pbmc <- FindVariableFeatures(pbmc, selection.method = "vst", nfeatures = 2000)
2
3# Identify the 10 most highly variable genes
4top10 <- head(VariableFeatures(pbmc), 10)
5
6# plot variable features with and without labels
7plot1 <- VariableFeaturePlot(pbmc)
8plot2 <- LabelPoints(plot = plot1, points = top10, repel = TRUE)
9
10CombinePlots(plots = list(plot1, plot2), ncol =1)
11
Scaling the data
ScaleData函数：
Shifts the expression of each gene, so that the mean expression across cells is 0
Scales the expression of each gene, so that the variance across cells is 1
我们一般将平均值为0，方差值为1的数据认为是标准数据
1all.genes <- rownames(pbmc)
2pbmc <- ScaleData(pbmc, features = all.genes)
如果数据量很大，这一步可能需要较长时间
在scale前后检查数据
1## 检查前后数据的区别
2#### raw counts, same as pbmc@assays$RNA@counts[1:6, 1:6]
3pbmc[["RNA"]]@counts[1:6, 1:6]
4### library size normalized and log transformed data
5pbmc[["RNA"]]@data[1:6, 1:6]
6### scaled data
7pbmc[["RNA"]]@scale.data[1:6, 1:6]
scale是Seurat工作流程中必不可少的一步。但结果仅限于用作PCA分析的输入。
ScaleData中默认设置是仅对先前标识的变量特征执行降维（默认为2000）.因此，在上一个函数调用中应省略features参数。
1pbmc <- ScaleData(pbmc, vars.to.regress = "percent.mt")
PCA
主成分分析（PCA）是一种线性降维技术
1pbmc <- RunPCA(pbmc, features = VariableFeatures(object = pbmc), verbose = FALSE)
2
3p1<- DimPlot(pbmc, reduction = "pca")
4p1
5
如果想了解更多PCA相关的，可以在YouTube观看StatQuest的: https://www.youtube.com/watch?v=HMOI_lkzW08
或者看下面的教程：
聚类分析和主成分分析
或者原作者的博客：
https://divingintogeneticsandgenomics.rbind.io/post/pca-in-action/
https://divingintogeneticsandgenomics.rbind.io/post/permute-test-for-pca-components/
当然你也可以用ggplot2画出各种好看的PCA图，网上搜索的话，画图代码有很多。这里不再论述。
确定PCs数
为了克服scRNA序列数据单一特征中的广泛技术噪音，Seurat根据其PCA分数对细胞进行聚类，每个PC基本上表示一个“元特征”，该特征结合了相关特征集上的信息。因此，最主要的主成分代表了数据集的强大压缩。但是，我们应该选择包括多少个PC？10个？20？还是100？
可以用如下方法来大致判定：
1pbmc <- JackStraw(pbmc, num.replicate = 100, dims = 50)
2pbmc <- ScoreJackStraw(pbmc, dims = 1:50)
3
4JackStrawPlot(pbmc, dims = 1:30)
5
1ElbowPlot(pbmc, ndims = 50)
variance explained by each PC
1mat <- pbmc[["RNA"]]@scale.data
2pca <- pbmc[["pca"]]
3
4# Get the total variance:
5total_variance <- sum(matrixStats::rowVars(mat))
6
7eigValues = (pca@stdev)^2 ## EigenValues
8varExplained = eigValues / total_variance
9
10varExplained %>% enframe(name = "PC", value = "varExplained" ) %>%
11 ggplot(aes(x = PC, y = varExplained)) +
12 geom_bar(stat = "identity") +
13 theme_classic() +
14 ggtitle("scree plot")
1### this is what Seurat is plotting: standard deviation
2pca@stdev %>% enframe(name = "PC", value = "Standard Deviation" ) %>%
3 ggplot(aes(x = PC, y = `Standard Deviation`)) +
4 geom_point() +
5 theme_classic()
细胞分群
1pbmc <- FindNeighbors(pbmc, dims = 1:20)
2pbmc <- FindClusters(pbmc, resolution = 0.5)
3# Look at cluster IDs of the first 5 cells
4head(Idents(pbmc), 5)
运行非线性降维(UMAP/tSNE)
1pbmc <- RunUMAP(pbmc, dims = 1:20)
2pbmc<- RunTSNE(pbmc, dims = 1:20)
3
4## after we run UMAP and TSNE, there are more entries in the reduction slot
5str(pbmc@reductions)
6
7DimPlot(pbmc, reduction = "umap", label = TRUE)
1## now let's visualize in the TSNE space
2DimPlot(pbmc, reduction = "tsne")
tSNE相关视频: https://www.youtube.com/watch?v=NEaUSP4YerM
1## now let's label the clusters in the PCA space
2DimPlot(pbmc, reduction = "pca")
查找差异表达特征（集群生物标记）
1# find all markers of cluster 1
2cluster1.markers <- FindMarkers(pbmc, ident.1 = 1, min.pct = 0.25)
3head(cluster1.markers, n = 5)
4# find all markers distinguishing cluster 5 from clusters 0 and 3
5cluster5.markers <- FindMarkers(pbmc, ident.1 = 5, ident.2 = c(0, 3), min.pct = 0.25)
6head(cluster5.markers, n = 5)
7# find markers for every cluster compared to all remaining cells, report only the positive ones
8pbmc.markers <- FindAllMarkers(pbmc, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)
9pbmc.markers %>% group_by(cluster) %>% top_n(n = 2, wt = avg_logFC)
10
这一步很费时间，如果你觉得慢，Seurat V3.0.2 为FindALLMarkers在内的一些步骤提供了并行支持。
更多了解：https://satijalab.org/seurat/v3.0/future_vignette.html
1# we only have 2 CPUs reserved for each one.
2plan("multiprocess", workers = 2)
3pbmc.markers <- FindAllMarkers(pbmc, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)
可视化marker基因
VlnPlot
1VlnPlot(pbmc, features = c("MS4A1", "CD79A"))
1## understanding the matrix of data slots
2pbmc[["RNA"]]@data[c("MS4A1", "CD79A"), 1:30]
3pbmc[["RNA"]]@scale.data[c("MS4A1", "CD79A"), 1:30]
4pbmc[["RNA"]]@counts[c("MS4A1", "CD79A"), 1:30]
5# you can plot raw counts as well
6VlnPlot(pbmc, features = c("MS4A1", "CD79A"), slot = "counts", log = TRUE)
1VlnPlot(pbmc, features = c("MS4A1", "CD79A"), slot = "scale.data")
FeaturePlot
plot the expression intensity overlaid on the Tsne/UMAP plot.
1FeaturePlot(pbmc, features = c("MS4A1", "GNLY", "CD3E", "CD14", "FCER1A", "FCGR3A", "LYZ", "PPBP", "CD8A"))
1p<- FeaturePlot(pbmc, features = "CD14")
2
3## before reordering
4p
1p_after<- p
2### after reordering
3p_after$data <- p_after$data[order(p_after$data$CD14),]
4
5CombinePlots(plots = list(p, p_after))
DoHeatmap
1top10 <- pbmc.markers %>% group_by(cluster) %>% top_n(n = 10, wt = avg_logFC)
2DoHeatmap(pbmc, features = top10$gene) + NoLegend()