单细胞代码解析-妇科癌症单细胞转录组及染色质可及性分析1：https://cloud.tencent.com/developer/article/2055573
单细胞代码解析-妇科癌症单细胞转录组及染色质可及性分析2：https://cloud.tencent.com/developer/article/2072069
单细胞代码解析-妇科癌症单细胞转录组及染色质可及性分析3：https://cloud.tencent.com/developer/article/2078159
图片.png###########################################################
# Part 3: scRNA-seq processing after doublet removal
###########################################################
setwd(dir)
# Load the RNA dataset
counts.init <- Read10X_h5(filename = "./filtered_feature_bc_matrix.h5")
# Initialize the Seurat object with the raw (non-normalized data).
rna <- CreateSeuratObject(counts = counts.init, min.cells = 3)# genes not present in at least 3 cells are removed
rna
rna <- rna[,cells]
#这部分放入了以前鉴定得到的双细胞的结果；%in%：意思都是x的值是否在y里面
rna <- rna[,!(colnames(rna) %in% doublets)]
dim(rna)
#Normalize；标准换
rna <- NormalizeData(rna, normalization.method = "LogNormalize", scale.factor = 10000)
#Feature Selection；寻找高变基因
rna <- FindVariableFeatures(rna, selection.method = "vst", nfeatures = 2000)
#Scaling
all.genes <- rownames(rna)
rna <- ScaleData(rna, features = all.genes)
rna <- RunPCA(rna)
feats <- list(stromal,immune,fibroblast,endothelial,epithelial,smooth,plasma)
#利用seurat包的一个打分函数AddModuleScore，依据基因的平均表达水平进行分析
rna <- AddModuleScore(rna,features = feats,name = c("stromal.","immune.","fibroblast.","endothelial.",
"epithelial.","smooth.","plasma."),search = T)
# Add PC1 to metadata
##对前面提到的细胞类群进行相关性分析
rna@meta.data$PC1 <- rna@reductions$pca@cell.embeddings[,1]
count_cor_PC1 <- cor(rna$PC1,rna$nCount_RNA,method = "spearman")
stromal.cor <- cor(rna$PC1,rna$stromal.1,method = "spearman")
immune.cor <- cor(rna$PC1,rna$immune.2,method = "spearman")
fibroblast.cor <- cor(rna$PC1,rna$fibroblast.3,method = "spearman")
endothelial.cor <- cor(rna$PC1,rna$endothelial.4,method = "spearman")
epithelial.cor <- cor(rna$PC1,rna$epithelial.5,method = "spearman")
smooth.cor <- cor(rna$PC1,rna$smooth.6,method = "spearman")
plasma.cor <- cor(rna$PC1,rna$plasma.7,method = "spearman")
# Make JackStraw Plot
rna <- JackStraw(rna, num.replicate = 100,dims = 50)
rna <- ScoreJackStraw(rna,dims = 1:50)
JackStrawPlot(rna, dims = 1:50) ggsave("JackStraw_postdoublet.png")
##下面是循环分析
# If PC1 is correalted with read depth, check to see if biological variation is corralted to PC1
#首先确定亚群分析的resolution确定，将resolution确定为0.7
#round：它四舍五入到给定的位数，如果没有提供四舍五入的位数，它会将数字四舍五入到最接近的整数
if (round(abs(count_cor_PC1),2) > 0.5){
if( round(abs(stromal.cor),2) >= 0.5 |
round(abs(immune.cor),2) >= 0.5 |
round(abs(fibroblast.cor),2) >= 0.5 |
round(abs(endothelial.cor),2) >= 0.5 |
round(abs(epithelial.cor),2) >= 0.5 |
round(abs(smooth.cor),2) >= 0.5 |
round(abs(plasma.cor),2) >= 0.5){
rna <- FindNeighbors(rna,dims = 1:50)
rna <- FindClusters(rna,resolution = 0.7)
rna <- RunUMAP(rna,dims = 1:50)
Idents(rna) <- "RNA_snn_res.0.7"
# Verify with inferCNV: is PC1 correlated with CNV events/Malignancy?
#########################################################################
# inferCNV: does PC1 also correlated with CNV/malignancy status?
#InferCNV是一个由broad研究所开发的，利用单细胞转录组数据分析肿瘤细胞拷贝数变异（CNV）的工具。基本思路是在整个基因组范围内，通过计算肿瘤细胞关于参考的“正常”细胞的相对表达，利用滑窗思想对邻近的基因相对表达，计算拷贝数。
library(infercnv)
library(stringr)
library(Seurat)
counts_matrix = GetAssayData(rna, slot="counts")
# Identify immune clusters
#######################################################
# Find immune cells by relative enrichment of ESTIMATE immune signature
#描述性统计 —— R 'psych' 表型数据的描述性统计，是对表型数据进行的基础分析，包括最大值、最小是、均值、方差、极差等
library(psych)
test <- VlnPlot(rna,features = "immune.2")
#describeBy：分组计算
data <- describeBy(test$data$immune.2, test$data$ident, mat = TRUE)
data.immune <- dplyr::filter(data,median > 0.1)
test <- VlnPlot(rna,features = "plasma.7")
data <- describeBy(test$data$plasma.7, test$data$ident, mat = TRUE)
data.plasma <- dplyr::filter(data,median > 0.1)
immune.clusters <- intersect(data.immune$group1,levels(Idents(rna)))
plasma.clusters <- intersect(data.plasma$group1,levels(Idents(rna)))
immune.clusters <- unique(append(immune.clusters,plasma.clusters))
for (i in 1:length(immune.clusters)){
j <- which(levels(Idents(rna)) == immune.clusters[i])
levels(Idents(rna))[j] <- paste0("immune.",immune.clusters[i])
}
rna@meta.data$postdoublet.idents <- Idents(rna)
idents <- data.frame(rownames(rna@meta.data),rna@meta.data$postdoublet.idents)
colnames(idents) <- c("V1","V2")
saveRDS(rna,"./rna_postdoublet_preinferCNV.rds")
# Make inferCNV inputs
rownames(idents) <- NULL
colnames(idents) <- NULL
write.table(idents,"./sample_annotation_file_inferCNV.txt",sep = "t",row.names = FALSE)
#R语言使用read.delim函数读取带分隔符的文本文件
##构建注释文件
idents <- read.delim("./sample_annotation_file_inferCNV.txt",header = F)
gtf <- read.delim(GRCH38.annotations,header = F)
library(EnsDb.Hsapiens.v86)
convert.symbol = function(Data){
ensembls <- Data$V1
ensembls <- gsub("\.[0-9]*$", "", ensembls)
geneIDs1 <- ensembldb::select(EnsDb.Hsapiens.v86, keys= ensembls, keytype = "GENEID", columns = "SYMBOL")
#cbind.fill：可填充缺失值并适用于任意数据类型
Data <- cbind.fill(Data, geneIDs1, fill = NA)
#去除NA
Data <- na.omit(Data)
Data$feature <- Data$SYMBOL
Data.new <- data.frame(Data$SYMBOL,Data$V2,Data$V3,Data$V4)
Data.new$Data.V2 <- paste("chr",Data.new$Data.V2,sep = "")
Data.new$Data.SYMBOL <- make.unique(Data.new$Data.SYMBOL)
return(Data.new)
}
##convert.symbol：基因id的转换
gtf <- convert.symbol(gtf)
head(gtf)
write.table(gtf,"./Homo_sapiens.GRCh38.86.symbol.txt",sep = "t",row.names = FALSE,col.names = FALSE)
#R 语言中的 length () 函数用于获取或设置向量 (列表)或其他对象的长度。
#CreateInfercnvObject () 函数构建infercnv对象，此处必须设置gene_order_file参数，其输入是一个基因的染色体位置信息文件，以制表符分隔
num.immune.clusters = length(immune.clusters)
# create the infercnv object
if ( num.immune.clusters == 1) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=paste0("immune.",immune.clusters[1]))
} else if (num.immune.clusters == 2) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2])))
} else if ( num.immune.clusters == 3 ) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3])))
} else if (num.immune.clusters == 4) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3]),
paste0("immune.",immune.clusters[4])))
} else if (num.immune.clusters == 5) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3]),
paste0("immune.",immune.clusters[4]),
paste0("immune.",immune.clusters[5])))
} else if (num.immune.clusters == 6) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3]),
paste0("immune.",immune.clusters[4]),
paste0("immune.",immune.clusters[5]),
paste0("immune.",immune.clusters[6])))
}else if (num.immune.clusters == 7) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3]),
paste0("immune.",immune.clusters[4]),
paste0("immune.",immune.clusters[5]),
paste0("immune.",immune.clusters[6]),
paste0("immune.",immune.clusters[7])))
}else if (num.immune.clusters == 8) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3]),
paste0("immune.",immune.clusters[4]),
paste0("immune.",immune.clusters[5]),
paste0("immune.",immune.clusters[6]),
paste0("immune.",immune.clusters[7]),
paste0("immune.",immune.clusters[8])))
}else if (num.immune.clusters == 9) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3]),
paste0("immune.",immune.clusters[4]),
paste0("immune.",immune.clusters[5]),
paste0("immune.",immune.clusters[6]),
paste0("immune.",immune.clusters[7]),
paste0("immune.",immune.clusters[8]),
paste0("immune.",immune.clusters[9])))
}else if (num.immune.clusters == 10) {
infercnv_obj = CreateInfercnvObject(raw_counts_matrix=as.matrix(counts_matrix),
annotations_file="./sample_annotation_file_inferCNV.txt",
gene_order_file="./Homo_sapiens.GRCh38.86.symbol.txt",
ref_group_names=c(paste0("immune.",immune.clusters[1]),
paste0("immune.",immune.clusters[2]),
paste0("immune.",immune.clusters[3]),
paste0("immune.",immune.clusters[4]),
paste0("immune.",immune.clusters[5]),
paste0("immune.",immune.clusters[6]),
paste0("immune.",immune.clusters[7]),
paste0("immune.",immune.clusters[8]),
paste0("immune.",immune.clusters[9]),
paste0("immune.",immune.clusters[10])))
}
#根据cluster的数量，进行else if的处理，去输出判断拷贝数的结果
# perform infercnv operations to reveal cnv signal
#infercnv::run()：两步法进行拷贝数分析，不需要一步一步进行尝试，参考来源：[https://www.jianshu.com/p/70f7a168fe62](https://www.jianshu.com/p/70f7a168fe62)
infercnv_obj = infercnv::run(infercnv_obj,
cutoff=0.1, # use 1 for smart-seq, 0.1 for 10x-genomics
out_dir="./output_dir_CNV_postdoublet_PassedPC1Checks", # dir is auto-created for storing outputs
cluster_by_groups=T, # cluster
denoise=T,scale_data = T,
HMM=T,HMM_type = "i6",analysis_mode = "samples",min_cells_per_gene = 10,
BayesMaxPNormal = 0.4, num_threads = 8
)
regions <- read.delim("./output_dir_CNV_postdoublet_PassedPC1Checks/HMM_CNV_predictions.HMMi6.hmm_mode-samples.Pnorm_0.4.pred_cnv_regions.dat")
probs <- read.delim("./output_dir_CNV_postdoublet_PassedPC1Checks/BayesNetOutput.HMMi6.hmm_mode-samples/CNV_State_Probabilities.dat")
#R 编程语言中的 as.data.frame() 函数用于将对象转换为数据框
probs <- as.data.frame(t(probs[3,]))
colnames(probs) <- "Prob.Normal"
probs <- dplyr::filter(probs,Prob.Normal < 0.05)
cnvs <- rownames(probs)
#gsub ()可以用于字段的删减、增补、替换和切割，可以处理一个字段也可以处理由字段组成的向量
cnvs <- gsub("\.","-",cnvs)
regions <- regions[regions$cnv_name %in% cnvs, ]
#sub R语言中的函数用于替换字符串中模式的第一个匹配项
cnv.groups <- sub("\..*", "", regions$cell_group_name)
length(which(rownames(rna@reductions$pca@cell.embeddings) == rownames(rna@meta.data)))
rna$PC1.loading <- rna@reductions$pca@cell.embeddings[,1]
rna$cell.barcode <- rownames(rna@meta.data)
#ifelse()中的条件判断中可以得到多个逻辑结果，有多少个逻辑结果，ifelse()的返回值就有多少个元素
rna$CNV.Pos <- ifelse(as.character(rna$postdoublet.idents) %in% cnv.groups,1,0)
cnv.freq <- data.frame(table(regions$cell_group_name))
cnv.freq$Var1 <- sub("\..*", "", cnv.freq$Var1)
rna$Total_CNVs <- ifelse(as.character(rna$postdoublet.idents) %in% cnv.freq$Var1,cnv.freq$Freq,0)
boxplot.cnv <- ggplot(rna@meta.data,aes(x= postdoublet.idents,y=PC1.loading,color = as.factor(CNV.Pos))) geom_boxplot()
boxplot.cnv ggsave("postdoublet_CNV_PC1_boxplot.png")
#describeBy：分组计算
data <- describeBy(boxplot.cnv$data$PC1.loading, boxplot.cnv$data$postdoublet.idents, mat = TRUE)
data$CNV <- ifelse(data$group1 %in% cnv.groups,1,0)
wilcox <- wilcox.test(data = rna@meta.data,PC1.loading~CNV.Pos)
##根据pvalue进行判断不同的数据属性
if (wilcox$p.value < 0.05){
rna <- rna
library(stringr)
#str_remove：删除字符串中的匹配模式。
levels(Idents(rna)) <- str_remove(levels(Idents(rna)),"immune.")
saveRDS(rna,"./rna_postdoublet_PassedPC1Checks.rds")
}else{
all.genes <- rownames(rna)
rna <- ScaleData(rna, features = all.genes,vars.to.regress = "nCount_RNA")
rna <- FindNeighbors(rna,dims = 1:50)
rna <- FindClusters(rna,resolution = 0.7)
rna <- RunUMAP(rna,dims = 1:50)
Idents(rna) <- "RNA_snn_res.0.7"
library(infercnv)
library(stringr)
library(Seurat)
counts_matrix = GetAssayData(rna, slot="counts")
总结
作者大大的代码也太长了吧，根本一篇都弄不完。上面分析的主要思路是前期通过对细胞类型鉴定，然后筛选出了双细胞结果，根据细胞类型进行下面的分析，这次加入了肿瘤变异之间的拷贝数分析，感觉自己的分析中也可以应用到这个内容。然后作者的代码里有一个提取注释文件的方法也很好，也是可以保留下来，以后提取注释文件用到的。还有作者经常用到的else if语句也很多，可以减少后面的不停调试的方法。
