# 4.7 实战七 | SMARTer | 人胰腺细胞

## 1 前言

这次使用的数据是：[Lawlor et al. (2017) ](https://pubmed.ncbi.nlm.nih.gov/27864352/)中的不同人类供体的胰腺细胞

SMARTer其实不是个像Fluidigm、10X一样的制备系统，它是一个试剂盒。SMART技术是Clontech（Takara旗下全资子公司）的专利技术，**2009年升级为SAMRTer技术后**，采用更灵敏的SMARTer Oligo和高效的SMARTScribe RT进行逆转录，使其灵敏度提高至皮克级。利用SMARTer技术**只需单管、单酶即可完成逆转录，无需接头连接**，减少了样品操作步骤，也极大降低了样品损失，保留了原始信息，为RNA-Seq提供了可靠的基础。

2013年 Clontech公司就为Fluidigm的C1单细胞全自动制备系统推出了SMARTer Ultra Low RNA Kit。当时也是能够从C1捕获的单细胞中产生mRNA-seq文库，方便了研究。Fluidigm在C1平台上检验了多款试剂盒，发现SMARTer cDNA合成是最可靠的方法之一

> 参考：<http://www.ebiotrade.com/newsf/2013-9/20139493414227.htm>

后来很多平台也在使用SMARTer试剂（如WaferGen ICELL8 Single-Cell System），不过后来发现Smart-seq2的扩增效果优于SMARTer试剂盒，而且Smart-seq2技术比传统的SMARTer方法能产生更长和更多的cDNA，在低表达量时，Smart-seq2比SMARTer的基因检出率更高，结果更稳定

### **数据准备**

```r
library(scRNAseq)
sce.lawlor <- LawlorPancreasData()
sce.lawlor
# class: SingleCellExperiment 
# dim: 26616 638 
# metadata(0):
#   assays(1): counts
# rownames(26616): ENSG00000229483 ENSG00000232849 ...
# ENSG00000251576 ENSG00000082898
# rowData names(0):
#   colnames(638): 10th_C10_S104 10th_C11_S96 ...
# 9th-C96_S81 9th-C9_S13
# colData names(8): title age ... race Sex
# reducedDimNames(0):
#   altExpNames(0):
```

### **ID转换**

```r
library(AnnotationHub)
edb <- AnnotationHub()[["AH73881"]]
anno <- select(edb, keys=rownames(sce.lawlor), keytype="GENEID", 
    columns=c("SYMBOL", "SEQNAME"))
rowData(sce.lawlor) <- anno[match(rownames(sce.lawlor), anno[,1]),-1]
rowData(sce.lawlor)
# DataFrame with 26616 rows and 2 columns
# SYMBOL     SEQNAME
# <character> <character>
#   ENSG00000229483   LINC00362          13
# ENSG00000232849   LINC00363          13
# ENSG00000229558    SACS-AS1          13
# ENSG00000232977   LINC00327          13
# ENSG00000227893   LINC00352          13
# ...                     ...         ...
# ENSG00000232746   LINC02022           3
# ENSG00000150867     PIP4K2A          10
# ENSG00000255021  AC093496.1           3
# ENSG00000251576   LINC01267           3
# ENSG00000082898        XPO1           2
```

## 2 质控

**依然是备份一下，把unfiltered数据主要用在质控的探索上**

```r
unfiltered <- sce.lawlor
```

**检查是否有线粒体基因和批次信息**

```r
# 其中有MT基因
table(rowData(sce.lawlor)$SEQNAME=="MT")
# 
# FALSE  TRUE 
# 25269    13 

# 还有一些批次信息
table(sce.lawlor$`islet unos id`)
# 
# ACCG268 ACCR015A ACEK420A  ACEL337  ACHY057  ACIB065  ACIW009  ACJV399 
# 136       57       45      103       39       57       93      108
```

**进行质控**

```r
stats <- perCellQCMetrics(sce.lawlor, 
    subsets=list(Mito=which(rowData(sce.lawlor)$SEQNAME=="MT")))
qc <- quickPerCellQC(stats, percent_subsets="subsets_Mito_percent",
    batch=sce.lawlor$`islet unos id`)
# 过滤了34个细胞
table(qc$discard)
# 
# FALSE  TRUE 
# 604    34 

sce.lawlor <- sce.lawlor[,!qc$discard]
```

**看看过滤掉多少**

```r
colSums(as.matrix(qc))
# low_lib_size            low_n_features high_subsets_Mito_percent                   discard 
# 9                         5                        25                        34
```

**作图看一下**

```r
colData(unfiltered) <- cbind(colData(unfiltered), stats)
unfiltered$discard <- qc$discard

gridExtra::grid.arrange(
    plotColData(unfiltered, x="islet unos id", y="sum", colour_by="discard") +
        scale_y_log10() + ggtitle("Total count") +
        theme(axis.text.x = element_text(angle = 90)),
    plotColData(unfiltered, x="islet unos id", y="detected", 
        colour_by="discard") + scale_y_log10() + ggtitle("Detected features") +
        theme(axis.text.x = element_text(angle = 90)), 
    plotColData(unfiltered, x="islet unos id", y="subsets_Mito_percent",
        colour_by="discard") + ggtitle("Mito percent") +
        theme(axis.text.x = element_text(angle = 90)),
    ncol=2
)
```

![](https://jieandze1314-1255603621.cos.ap-guangzhou.myqcloud.com/blog/2020-07-20-112030.png)

**看一下文库大小和线粒体占比的关系**

```r
plotColData(unfiltered, x="sum", y="subsets_Mito_percent",
    colour_by="discard") + scale_x_log10()
```

![](https://jieandze1314-1255603621.cos.ap-guangzhou.myqcloud.com/blog/2020-07-20-112119.png)

**最后把过滤条件应用在原数据**

```r
sce.lawlor <- sce.lawlor[,!qc$discard]
```

## 3 归一化

继续使用去卷积方法

```r
library(scran)
set.seed(1000)
clusters <- quickCluster(sce.lawlor)
sce.lawlor <- computeSumFactors(sce.lawlor, clusters=clusters)
sce.lawlor <- logNormCounts(sce.lawlor)

summary(sizeFactors(sce.lawlor))
# Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
# 0.2955  0.7807  0.9633  1.0000  1.1820  2.6287
```

## 4 找表达量高变化基因

这里没有ERCC也没有UMI，所以就用最基础的方法构建模型：`modelGeneVar`

不过还是要指定批次信息

```r
dec.lawlor <- modelGeneVar(sce.lawlor, block=sce.lawlor$`islet unos id`)
chosen.genes <- getTopHVGs(dec.lawlor, n=2000)
```

## 5 【尝试】矫正批次

这里写“尝试”是因为这里有一个问题：**细胞总数不多，才600个，但批次的数量很多**，所以归到单独的批次上细胞数就很少。这时如果继续矫正批次，**不知道会不会抹除一些真实的生物学特性。**

归根结底，还是一个技术噪音与生物因素之间的取舍问题

```r
table(sce.lawlor$`islet unos id`)
# 
# ACCG268 ACCR015A ACEK420A  ACEL337  ACHY057  ACIB065  ACIW009  ACJV399 
# 136       57       45      103       39       57       93      108
```

所以可以尝试一下：

```r
library(batchelor)
set.seed(1001010)
merged.lawlor <- fastMNN(sce.lawlor, subset.row=chosen.genes, 
                         batch=sce.lawlor$`islet unos id`)

metadata(merged.lawlor)$merge.info$lost.var
```

关于这个结果：**`lost.var` ，值越大表示丢失的真实生物异质性越多**

* It contains a matrix of the **variance lost in each batch** (column) at each merge step (row).
* Large proportions of lost variance **(>10%)** suggest that correction is **removing genuine biological heterogeneity.**&#x20;

![](https://jieandze1314-1255603621.cos.ap-guangzhou.myqcloud.com/blog/2020-07-20-113144.png)

看到的确有损失生物异质性的可能性，**那么就先放弃这个计划，直接进行下面的降维**

## 5 降维聚类

### **降维**

```r
library(BiocSingular)
set.seed(101011001)
sce.lawlor <- runPCA(sce.lawlor, subset_row=chosen.genes, ncomponents=25)
sce.lawlor <- runTSNE(sce.lawlor, dimred="PCA")
```

### **聚类**

```r
snn.gr <- buildSNNGraph(sce.lawlor, use.dimred="PCA")
colLabels(sce.lawlor) <- factor(igraph::cluster_walktrap(snn.gr)$membership)
```

**看分群与细胞类型之间关系**

```r
tab <- table(colLabels(sce.lawlor), sce.lawlor$`cell type`)
library(pheatmap)
pheatmap(log10(tab+10), color=viridis::viridis(100))
```

![](https://jieandze1314-1255603621.cos.ap-guangzhou.myqcloud.com/blog/2020-07-20-113541.png)

**看分群与批次之间**

```r
tab2 <- table(colLabels(sce.lawlor), sce.lawlor$`islet unos id`)
library(pheatmap)
pheatmap(log10(tab2+10), color=viridis::viridis(100))
```

![](https://jieandze1314-1255603621.cos.ap-guangzhou.myqcloud.com/blog/2020-07-20-113638.png)

### **最后看看批次效应**

```r
gridExtra::grid.arrange(
    plotTSNE(sce.lawlor, colour_by="label"),
    plotTSNE(sce.lawlor, colour_by="islet unos id"),
    ncol=2
)
```

![](https://jieandze1314-1255603621.cos.ap-guangzhou.myqcloud.com/blog/2020-07-20-113811.png)