生信技能树单细胞数据挖掘笔记(2)

生信技能树单细胞数据挖掘笔记(1)：数据读入

生信技能树单细胞数据挖掘笔记(2)：创建Seurat对象并进行质控、筛选高变基因并可视化
生信技能树单细胞数据挖掘笔记(3)：降维与聚类

生信技能树单细胞数据挖掘笔记(4)：其他分析（周期判断、double诊断、细胞类型注释）
生信技能树单细胞数据挖掘笔记(5)：轨迹分析

创建Seurat对象并质控

创建Seurat对象，然后过滤掉表达量过低的基因、表达基因过少的细胞以及线粒体基因过多的细胞

### 2、构建seurat对象，质控绘图----
# 2.1 构建seurat对象，质控
#In total, 2,343 cells from tumor cores were included in this analysis.
#quality controlstandards: 
#1) genes detected in < 3 cells were excluded; 筛选基因
#2) cells with < 50 total detected genes were excluded; 筛选细胞 
#3) cells with ≥ 5% of mitochondria-expressed genes were excluded. 筛选细胞
library("Seurat")
sce.meta <- data.frame(Patient_ID=group$Patient_ID,
                       row.names = group$sample)
head(sce.meta)
table(sce.meta$Patient_ID)
# 这个函数 CreateSeuratObject 有多种多样的执行方式
scRNA = CreateSeuratObject(counts=a.filt,
                           meta.data = sce.meta,
                           min.cells = 3, 
                           min.features = 50)
#counts:a matrix-like object with unnormalized data with cells as columns and features as rows 
#meta.data:Additional cell-level metadata to add to the Seurat object
#min.cells: features detected in at least this many cells. 
#min.features:cells where at least this many features are detected.
head([email protected])
#nCount_RNA：the number of cell total counts
#nFeature_RNA：the number of cell's detected gene
summary([email protected])
scRNA@assays$RNA@counts[1:4,1:4]
# 可以看到，之前的counts矩阵存储格式发生了变化：4 x 4 sparse Matrix of class "dgCMatrix"

dim(scRNA)
#  20047  2342 仅过滤掉一个细胞
#接下来根据线粒体基因表达筛选低质量细胞
#Calculate the proportion of transcripts mapping to mitochondrial genes
table(grepl("^MT-",rownames(scRNA)))
#FALSE 


#20050 没有染色体基因
scRNA[["percent.mt"]] <- PercentageFeatureSet(scRNA, pattern = "^MT-")
head([email protected])
summary([email protected])
#结果显示没有线粒体基因，因此这里过滤也就没有意义，但是代码留在这里
# 万一大家的数据里面有线粒体基因，就可以如此这般进行过滤啦。
pctMT=5 #≥ 5% of mitochondria-expressed genes
scRNA <- subset(scRNA, subset = percent.mt < pctMT)
dim(scRNA)


table(grepl("^ERCC-",rownames(scRNA)))
#FALSE  TRUE 
#19961    86  发现是有ERCC基因
#External RNA Control Consortium，是常见的已知浓度的外源RNA分子spike-in的一种
#指标含义类似线粒体含量，ERCC含量大，则说明total sum变小
scRNA[["percent.ERCC"]] <- PercentageFeatureSet(scRNA, pattern = "^ERCC-")
head([email protected])
summary([email protected])
rownames(scRNA)[grep("^ERCC-",rownames(scRNA))]
[1] "ERCC-00002" "ERCC-00003" "ERCC-00004" "ERCC-00009" "ERCC-00012" "ERCC-00013" "ERCC-00014" "ERCC-00017" "ERCC-00019" "ERCC-00022" "ERCC-00024" "ERCC-00025"
[13] "ERCC-00028" "ERCC-00031" "ERCC-00033" "ERCC-00034" "ERCC-00035" "ERCC-00039" "ERCC-00040" "ERCC-00041" "ERCC-00042" "ERCC-00043" "ERCC-00044" "ERCC-00046"
[25] "ERCC-00051" "ERCC-00053" "ERCC-00054" "ERCC-00058" "ERCC-00059" "ERCC-00060" "ERCC-00062" "ERCC-00067" "ERCC-00069" "ERCC-00071" "ERCC-00073" "ERCC-00074"
[37] "ERCC-00076" "ERCC-00077" "ERCC-00078" "ERCC-00079" "ERCC-00081" "ERCC-00083" "ERCC-00084" "ERCC-00085" "ERCC-00086" "ERCC-00092" "ERCC-00095" "ERCC-00096"
[49] "ERCC-00097" "ERCC-00098" "ERCC-00099" "ERCC-00104" "ERCC-00108" "ERCC-00109" "ERCC-00111" "ERCC-00112" "ERCC-00113" "ERCC-00116" "ERCC-00120" "ERCC-00123"
[61] "ERCC-00126" "ERCC-00130" "ERCC-00131" "ERCC-00134" "ERCC-00136" "ERCC-00137" "ERCC-00138" "ERCC-00142" "ERCC-00143" "ERCC-00144" "ERCC-00145" "ERCC-00147"
[73] "ERCC-00148" "ERCC-00150" "ERCC-00154" "ERCC-00156" "ERCC-00157" "ERCC-00158" "ERCC-00160" "ERCC-00162" "ERCC-00163" "ERCC-00164" "ERCC-00165" "ERCC-00168"
[85] "ERCC-00170" "ERCC-00171"
#可以看到有不少ERCC基因

sum(scRNA$percent.ERCC< 40)
#较接近原文过滤数量2149，但感觉条件有点宽松了，先做下去看看
#网上看了相关教程，一般ERCC占比不高于10%
sum(scRNA$percent.ERCC< 10)   #就只剩下460个cell，明显低于文献中的数量
pctERCC=40
scRNA <- subset(scRNA, subset = percent.ERCC < pctERCC)
dim(scRNA)
# 20047  2142   原文为19752  2149
dim(a.filt)
#23460  2343 未过滤前

质控绘图

# 2.2 可视化
#图A：观察不同组cell的counts、feature分布
col.num <- length(unique([email protected]$Patient_ID))
library(ggplot2)

p1_1.1 <- VlnPlot(scRNA,
                features = c("nFeature_RNA"),
                group.by = "Patient_ID",
                cols =rainbow(col.num)) +
  theme(legend.position = "none") +
  labs(tag = "A")
p1_1.1
p1_1.2 <- VlnPlot(scRNA,
                features = c("nCount_RNA"),
                group.by = "Patient_ID",
                cols =rainbow(col.num)) +
  theme(legend.position = "none") 
p1_1.2
p1_1 <- p1_1.1 | p1_1.2
p1_1
VlnPlot(scRNA,
        features = c("nFeature_RNA","nCount_RNA","percent.ERCC"))
#图B：nCount_RNA与对应的nFeature_RNA关系
p1_2 <- FeatureScatter(scRNA, feature1 = "nCount_RNA", feature2 = "nFeature_RNA",
                       group.by = "Patient_ID",pt.size = 1.3) +
  labs(tag = "B")
p1_2
FeatureScatter(scRNA, feature1 = "nCount_RNA", feature2 = "percent.ERCC")

p1_1

p1_2

目前已经完成了文献中的fig1A、B

挑选高变基因

### 3、挑选hvg基因，可视化----
#highly Variable gene:简单理解sd大的
scRNA <- FindVariableFeatures(scRNA, selection.method = "vst", nfeatures = 1500) 
#根据文献原图，挑选变化最大的1500个hvg
top10 <- head(VariableFeatures(scRNA), 10) 
top10
plot1 <- VariableFeaturePlot(scRNA) 
#标记top10 hvg
p1_3 <- LabelPoints(plot = plot1, points = top10, repel = TRUE, size=2.5) +
  theme(legend.position = c(0.1,0.8)) +
  labs(tag = "C")
p1_3
save(scRNA, file="2.3.Rdata")

p1_3

目前已复现文献中的图1A到C。

---

转载请注明：周小钊的博客>>>生信技能树单细胞数据挖掘笔记(2)