跟着大神学单细胞数据分析

2020 年 3 月 30 日
笔记

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

wget http://cf.10xgenomics.com/samples/cell-exp/3.0.2/5k_pbmc_v3/5k_pbmc_v3_filtered_feature_bc_matrix.tar.gz    tar xvzf 5k_pbmc_v3_filtered_feature_bc_matrix.tar.gz

安装所需的R包

install.packages("tidyverse")  install.packages("rmarkdown")  install.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

# 读取PBMC数据集  pbmc.data <- Read10X(data.dir = "filtered_feature_bc_matrix/")  # 使用原始数据（未归一化处理）初始化Seurat对象  pbmc <- CreateSeuratObject(counts = pbmc.data, project = "pbmc5k", min.cells = 3, min.features = 200)  pbmc

An object of class Seurat  18791 features across 4962 samples within 1 assay  Active 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

## 使用Srurat V2 创建对象  pbmc <- CreateSeuratObject(raw.data = pbmc.data, project = "pbmc5k", min.cells = 3, min.features = 200)    pbmc    An object of class seurat in project pbmc5k   18791 genes across 5025 samples.

质量控制

## check at metadata  head([email protected])  # The [[ operator can add columns to object metadata. This is a great place to stash QC stats  pbmc[["percent.mt"]] <- PercentageFeatureSet(pbmc, pattern = "^MT-")  [email protected] %>% head()    ##将质量控制指标可视化为小提琴图  VlnPlot(pbmc, features = c("nFeature_RNA", "nCount_RNA", "percent.mt"), ncol = 3)    #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。  pbmc <- 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"

pbmc <- NormalizeData(pbmc, normalization.method = "LogNormalize", scale.factor = 10000)

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

pbmc <- FindVariableFeatures(pbmc, selection.method = "vst", nfeatures = 2000)    # Identify the 10 most highly variable genes  top10 <- head(VariableFeatures(pbmc), 10)    # plot variable features with and without labels  plot1 <- VariableFeaturePlot(pbmc)  plot2 <- LabelPoints(plot = plot1, points = top10, repel = TRUE)    CombinePlots(plots = list(plot1, plot2), ncol =1)

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

all.genes <- rownames(pbmc)  pbmc <- 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

## 检查前后数据的区别  #### raw counts, same as pbmc@assays$RNA@counts[1:6, 1:6]  pbmc[["RNA"]]@counts[1:6, 1:6]  ### library size normalized and log transformed data  pbmc[["RNA"]]@data[1:6, 1:6]  ### scaled data  pbmc[["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。

pbmc <- ScaleData(pbmc, vars.to.regress = "percent.mt")

PCA

主成分分析（PCA）是一种线性降维技术

pbmc <- RunPCA(pbmc, features = VariableFeatures(object = pbmc), verbose = FALSE)    p1<- DimPlot(pbmc, reduction = "pca")  p1

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：

pbmc <- JackStraw(pbmc, num.replicate = 100, dims = 50)  pbmc <- ScoreJackStraw(pbmc, dims = 1:50)    JackStrawPlot(pbmc, dims = 1:30)

ElbowPlot(pbmc, ndims = 50)

variance explained by each PC

mat <- pbmc[["RNA"]]@scale.data  pca <- pbmc[["pca"]]    # Get the total variance:  total_variance <- sum(matrixStats::rowVars(mat))    eigValues = (pca@stdev)^2  ## EigenValues  varExplained = eigValues / total_variance    varExplained %>% enframe(name = "PC", value = "varExplained" ) %>%    ggplot(aes(x = PC, y = varExplained)) +    geom_bar(stat = "identity") +    theme_classic() +    ggtitle("scree plot")

### this is what Seurat is plotting: standard deviation  pca@stdev %>% enframe(name = "PC", value = "Standard Deviation" ) %>%    ggplot(aes(x = PC, y = `Standard Deviation`)) +    geom_point() +    theme_classic()

细胞分群

pbmc <- FindNeighbors(pbmc, dims = 1:20)  pbmc <- FindClusters(pbmc, resolution = 0.5)  # Look at cluster IDs of the first 5 cells  head(Idents(pbmc), 5)

运行非线性降维(UMAP/tSNE)

pbmc <- RunUMAP(pbmc, dims = 1:20)  pbmc<- RunTSNE(pbmc, dims = 1:20)    ## after we run UMAP and TSNE, there are more entries in the reduction slot  str(pbmc@reductions)    DimPlot(pbmc, reduction = "umap", label = TRUE)

## now let's visualize in the TSNE space  DimPlot(pbmc, reduction = "tsne")

tSNE相关视频: https://www.youtube.com/watch?v=NEaUSP4YerM

## now let's label the clusters in the PCA space  DimPlot(pbmc, reduction = "pca")

查找差异表达特征（集群生物标记）

# find all markers of cluster 1  cluster1.markers <- FindMarkers(pbmc, ident.1 = 1, min.pct = 0.25)  head(cluster1.markers, n = 5)  # find all markers distinguishing cluster 5 from clusters 0 and 3  cluster5.markers <- FindMarkers(pbmc, ident.1 = 5, ident.2 = c(0, 3), min.pct = 0.25)  head(cluster5.markers, n = 5)  # find markers for every cluster compared to all remaining cells, report only the positive ones  pbmc.markers <- FindAllMarkers(pbmc, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)  pbmc.markers %>% group_by(cluster) %>% top_n(n = 2, wt = avg_logFC)

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

# we only have 2 CPUs reserved for each one.  plan("multiprocess", workers = 2)  pbmc.markers <- FindAllMarkers(pbmc, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25)

可视化marker基因

VlnPlot

VlnPlot(pbmc, features = c("MS4A1", "CD79A"))

## understanding the matrix of data slots  pbmc[["RNA"]]@data[c("MS4A1", "CD79A"), 1:30]  pbmc[["RNA"]]@scale.data[c("MS4A1", "CD79A"), 1:30]  pbmc[["RNA"]]@counts[c("MS4A1", "CD79A"), 1:30]  # you can plot raw counts as well  VlnPlot(pbmc, features = c("MS4A1", "CD79A"), slot = "counts", log = TRUE)

VlnPlot(pbmc, features = c("MS4A1", "CD79A"), slot = "scale.data")

FeaturePlot plot the expression intensity overlaid on the Tsne/UMAP plot.

FeaturePlot(pbmc, features = c("MS4A1", "GNLY", "CD3E", "CD14", "FCER1A", "FCGR3A", "LYZ", "PPBP", "CD8A"))

p<- FeaturePlot(pbmc, features = "CD14")    ## before reordering  p

p_after<- p  ### after reordering  p_after$data <- p_after$data[order(p_after$data$CD14),]    CombinePlots(plots = list(p, p_after))

DoHeatmap

top10 <- pbmc.markers %>% group_by(cluster) %>% top_n(n = 10, wt = avg_logFC)  DoHeatmap(pbmc, features = top10$gene) + NoLegend()