Seurat-多样本整合(rPCA)
R版本与运行环境信息
Date:2021-7-21sessionInfo("Seurat")R version 4.0.3 (2020-10-10)Platform: x86_64-w64-mingw32/x64 (64-bit)Running under: Windows 10 x64 (build 18363)version: Seurat_4.0.1
Tim S, Andrew Butler, Paul Hoffman , et al. Comprehensive integration of single cell data[J].Cell,2019.
数据整合原理
见Seurat-多样本整合(CCA)
https://www.jianshu.com/p/32ca61450fe9
通常一下情况可以选择使用rPCA的方法对数据进行整合
- 一个数据集中有大部分的cell无法与另一个数据集进行匹配( substantial fraction of cells in one dataset have no matching type in the other)
- 数据集均来自同一平台，如10X( Datasets originate from the same platform)
- 数据集非常的大(There are a large number of datasets or cells to integrate)
与基于CCA的数据整合的区别
- 在确定两个数据集共享的变异来源时，CCA更适合细胞类型比较保守，但是基因表达差异很大的数据，但是CCA可能会造成过度矫正，尤其是在非常大的数据集已经共享细胞较少的数据集
By identifying shared sources of variation between datasets, CCA is well-suited for identifying anchors when cell types are conserved, but there are very substantial differences in gene expression across experiments. CCA-based integration therefore enables integrative analysis when experimental conditions or disease states introduce very strong expression shifts, or when integrating datasets across modalities and species. However, CCA-based integration may also lead to overcorrection, especially when a large proportion of cells are non-overlapping across datasets.
- rPCA是一种运行速度更快的方法，同时也更为保守，因此整合后的数据可能并没有那么“整齐”，同时在执行rPCA寻找锚点时，需要先对每个需要整合的数据运行PCA分析，因此rPCA的方法更适合上述三种情况
RPCA-based integration runs significantly faster, and also represents a more conservative approach where cells in different biological states are less likely to ‘align’ after integration.
使用rPCA的方法整合数据
载入数据与相关包
rm(list = ls())library(SeuratData)library(Seurat)library(tidyverse)# install dataset#InstallData("ifnb")#载入ifnb数据LoadData("ifnb")
分割对象并对每组数据进行处理
#分割对象ifnb.list <- SplitObject(ifnb,split.by = "stim")#———查看数据----> ifnb.list$CTRLAn object of class Seurat14053 features across 6548 samples within 1 assayActive assay: RNA (14053 features, 0 variable features)$STIMAn object of class Seurat14053 features across 7451 samples within 1 assayActive assay: RNA (14053 features, 0 variable features)#————————----#归一化，寻找高变基因ifnb.list <- lapply(ifnb.list, function(x){x <- NormalizeData(x)x <- FindVariableFeatures(x,selection.method = "vst",nfeatures = 2000)})#确定用于数据整合的featuresfeatures <- SelectIntegrationFeatures(ifnb.list)#数据标准化和PCA分析ifnb.list <- lapply(ifnb.list,function(x){x <- ScaleData(x,features = features)x <- RunPCA(x,features = features)})
使用rPCA确定锚点并整合数据
#确定锚点ifnb.imm <- FindIntegrationAnchors(object.list = ifnb.list,anchor.features = features,reduction = "rpca")#整合数据ifnb.immdata <- IntegrateData(ifnb.imm)
整合数据标准分析
#设定分析用的数据集DefaultAssay(ifnb.immdata) <- "integrated"#标准分析ifnb.immdata <- ScaleData(ifnb.immdata, verbose = FALSE)ifnb.immdata <- RunPCA(ifnb.immdata, npcs = 30, verbose = FALSE)ifnb.immdata <- RunUMAP(ifnb.immdata, reduction = "pca", dims = 1:30)ifnb.immdata <- FindNeighbors(ifnb.immdata, reduction = "pca", dims = 1:30)ifnb.immdata <- FindClusters(ifnb.immdata, resolution = 0.5)p1 <- DimPlot(ifnb.immdata, reduction = "umap", group.by = "stim")p2 <- DimPlot(ifnb.immdata, reduction = "umap", group.by = "seurat_annotations", label = TRUE,repel = TRUE)p1 + p2
从上图可以看出，与CCA的整合相比，rPCA的整合结果要略差于CCA，但是可以通过调整FindIntegrationAnchors()的参数进一步优化整合结果
优化rPCA寻找锚点的结果
通过调整k.anchor参数, 默认为5
**k.anchor**： How many neighbors (k) to use when picking anchors, 5
#调整k.anchorifnb.imm <- FindIntegrationAnchors(object.list = ifnb.list,anchor.features = features,reduction = "rpca",k.anchor = 20)ifnb.immdata <- IntegrateData(ifnb.imm)DefaultAssay(ifnb.immdata) <- "integrated"# Run the standard workflow for visualization and clusteringifnb.immdata <- ScaleData(ifnb.immdata, verbose = FALSE)ifnb.immdata <- RunPCA(ifnb.immdata, npcs = 30, verbose = FALSE)ifnb.immdata <- RunUMAP(ifnb.immdata, reduction = "pca", dims = 1:30)ifnb.immdata <- FindNeighbors(ifnb.immdata, reduction = "pca", dims = 1:30)ifnb.immdata <- FindClusters(ifnb.immdata, resolution = 0.5)p1 <- DimPlot(ifnb.immdata, reduction = "umap", group.by = "stim")p2 <- DimPlot(ifnb.immdata, reduction = "umap", group.by = "seurat_annotations", label = TRUE,repel = TRUE)p1 + p2
可以看出整合的效果要好于默认参数
使用SCTransform对数据标准化
上服务器
NormalizeData()、FindVariableFeatures()和ScaleData()被SCTransform替代
library(glmGamPoi)LoadData("ifnb")ifnb.list <- SplitObject(ifnb, split.by = "stim")#执行SCTransform标准化ifnb.list <- lapply(X = ifnb.list, FUN = SCTransform, method = "glmGamPoi")#选择用于整合的featuresfeatures <- SelectIntegrationFeatures(object.list = ifnb.list, nfeatures = 3000)#Prepare an object list normalized with sctransform for integration.?PrepSCTIntegration#PCA分析ifnb.list <- lapply(X = ifnb.list, FUN = RunPCA, features = features)
数据整合
- 在常规分析中，使用少量的PC既能关注到关键的生物学差异，又能够不引入更多的技术差异，相当于一种保守性的做法。是的，它会失去一些生物差异信息，但是同时又在常规手段中比较安全。
- 这里使用的
sctransform，显然更“自信”一些，能提取更多的生物差异，并且兼顾不引入技术误差
常规分析中的FindVariableFeatures默认得到2000个高变异基因（HVGs），而这里的**sctransform**因为使用了更多的PCs，算法也更优化，所以默认会得到3000个HVGs。sctransform认为：新增加的这1000个基因就包含了之前没有检测到的微弱的生物学差异。而且，即使使用全部的全部的基因去做下游分析，得到的结果也是和sctransform这3000个基因的结果相似
#确定锚点immune.anchors <- FindIntegrationAnchors(object.list = ifnb.list,normalization.method = "SCT",anchor.features = features,dims = 1:30,reduction = "rpca",k.anchor = 5)#数据整合immune.combined.sct <- IntegrateData(anchorset = immune.anchors, normalization.method = "SCT", dims = 1:30)
整合后的数据UMAP分析
immune.combined.sct <- RunPCA(immune.combined.sct, verbose = FALSE)immune.combined.sct <- RunUMAP(immune.combined.sct, reduction = "pca", dims = 1:30)p1 <- DimPlot(immune.combined.sct, reduction = "umap", group.by = "stim")p2 <- DimPlot(immune.combined.sct, reduction = "umap", group.by = "seurat_annotations", label = TRUE,repel = TRUE)cowplot::plot_grid(p1,p2,nrow = 2)
若有收获，就点个赞吧
0 人点赞
