---
title: 使用SCTransform标准化
tags: []
id: '2358'
categories:
  - - 预处理
date: 2022-09-29 00:24:45
---

## 安装包

*   conda create -n seurat -c conda-forge r-seurat=4.1.1 -y
*   conda activate seurat
*   conda install -c conda-forge r-tidyverse -y
*   conda install -c conda-forge r-irkernel -y
*   Rscript -e "IRkernel::installspec(name='seurat', displayname='r-seurat')"
*   conda install -c conda-forge r-biocmanager -y
*   \# conda install -c conda-forge r-sctransform -y
*   \# conda install -c conda-forge r-rocr -y
*   conda install -c conda-forge r-modes -y
*   \# conda install -c conda-forge r-kernsmooth -y
*   conda install -c conda-forge r-fields -y
*   conda install -c conda-forge r-devtools -y
*   conda install -c conda-forge r-clustree -y
*   BiocManager::install("glmGamPoi")
*   ~/dev/xray/xray -c ~/etc/xui2.json &
*   wget -e "https\_proxy=http://127.0.0.1:20809" https://github.com/chris-mcginnis-ucsf/DoubletFinder/archive/refs/heads/master.zip -O DoubletFinder-master.zip
*   devtools::install\_local('DoubletFinder-master.zip')

## 读取数据

### 从UMI矩阵

```R
sce <- read.table(gzfile("GSM4203181_data.matrix.txt.gz"), 
                  header = T, row.names = 1)
sce <- Seurat::CreateSeuratObject(sce, project = 'GSM4203181',
                            min.cells = 3, min.features = 200)
sce[["percent.mt"]] <- Seurat::PercentageFeatureSet(sce, pattern = "^MT-")
sce[["percent.ERCC"]] <- Seurat::PercentageFeatureSet(sce, pattern = "^ERCC-")
tmp <- rownames(sce@meta.data)
sce[["orig.ident"]] <- paste0('sample', substr(tmp,18,nchar(tmp)))
table(sce[["orig.ident"]])
sce@meta.data
saveRDS(sce, 'GSM4203181.rds')
```

### cellranger输出

```R
sce <- Seurat::Read10X('filtered_feature_bc_matrix')
sce <- Seurat::CreateSeuratObject(sce, project = 'SRX6887740',
                            min.cells = 3, min.features = 200)
sce[["percent.mt"]] <- Seurat::PercentageFeatureSet(sce, pattern = "^MT-")
sce[["percent.ERCC"]] <- Seurat::PercentageFeatureSet(sce, pattern = "^ERCC-")
```

## 初步QC

```R
options(repr.plot.width = 12, repr.plot.height = 6)
Seurat::VlnPlot(sce, features = c("nFeature_RNA", "nCount_RNA", "percent.mt"), ncol = 3)
plot1 <- Seurat::FeatureScatter(sce, feature1 = "nCount_RNA", feature2 = "percent.mt")
plot2 <- Seurat::FeatureScatter(sce, feature1 = "nCount_RNA", feature2 = "nFeature_RNA")
plot1 + plot2
sce <- subset(sce, nFeature_RNA > 2500 & nFeature_RNA < 7500 & nCount_RNA > 2500 & nCount_RNA < 50000 & percent.mt < 7.5)
```

### 计算细胞周期评分

```R
sce <- Seurat::NormalizeData(sce)
g2m_genes <- Seurat::CaseMatch(search=Seurat::cc.genes$g2m.genes, 
                               match=rownames(sce))
s_genes <- Seurat::CaseMatch(search=Seurat::cc.genes$s.genes, 
                             match=rownames(sce))
sce <- Seurat::CellCycleScoring(sce, g2m.features=g2m_genes, s.features=s_genes)
sce$CC.Difference <- sce$S.Score - sce$G2M.Score
```

## SCTransform标准化

```R
sce <- Seurat::SCTransform(sce, vst.flavor = "v2",
                           vars.to.regress = c("CC.Difference", "percent.mt"),
                           verbose = F)
```

## 查看分群走向

### 降维

```R
sce <- Seurat::RunPCA(sce, assay="SCT", verbose = FALSE)
Seurat::ElbowPlot(sce, ndims = 50)
sce <- Seurat::RunUMAP(sce, reduction = "pca", 
                       dims = 1:30, verbose = FALSE)
sce <- Seurat::FindNeighbors(sce, reduction = "pca", 
                             dims = 1:30, verbose = FALSE)
```

### 分群

```R
sce <- Seurat::FindClusters(
    object = sce,
    resolution = c(seq(.1,1.6,.2)) #起始粒度，结束粒度，间隔
)
options(repr.plot.width = 12, repr.plot.height = 12)
require(clustree)
clustree::clustree(sce@meta.data, prefix = "SCT_snn_res.")
```

![](https://img-cdn.limour.top/2022/09/28/63346be4bd300.png)

这图还挺好看的

## 使用DoubletFinder去除doublet

### pK Identification

```R
sweep.res <- DoubletFinder::paramSweep_v3(sce, PCs = 1:10, sct = T, num.cores=8)
sweep.stats <- DoubletFinder::summarizeSweep(sweep.res, GT = FALSE)
bcmvn <- DoubletFinder::find.pK(sweep.stats)
```

### Run DoubletFinder

![](https://img-cdn.limour.top/2022/09/29/633471b0dea83.png)

**[10X genomics单细胞平台的dulblets比率](https://cloud.tencent.com/developer/article/1825672)**

```R
## Assuming 1.6% doublet formation rate - tailor for your dataset
nExp_poi <- round(0.016*nrow(sce@meta.data))
pK_bcmvn <- as.numeric(as.character(bcmvn$pK[which.max(bcmvn$BCmetric)]))
sce <- DoubletFinder::doubletFinder_v3(sce, PCs = 1:10, 
                                       pN = 0.25, pK = pK_bcmvn, 
                                       nExp = nExp_poi, reuse.pANN = FALSE, 
                                       sct = T)
sce@meta.data
DimPlot(sce, reduction = "umap", group.by = c("DF.classifications_0.25_0.2_20"))
```

```R
sce <- subset(sce, DF.classifications_0.25_0.2_20 != 'Doublet')
saveRDS(sce, 'SRX6887740.rds')
```