本文是参考学习CNS图表复现09—上皮细胞可以区分为恶性与否的学习笔记。可能根据学习情况有所改动。
文章的第一次分群拿到了epithelial cells [n = 5,581], 我们的CNS图表复现拿到的上皮细胞数量是：
> table([email protected]$immune_annotation)
epi immune stromal
5444 13792 4278
我在单细胞天地的教程：是否是免疫细胞很容易区分那是否是肿瘤细胞呢？提到过，常规做法是使用inferCNV算法可以区分细胞恶性与否。比如online 29 April 2020的文章《Single-Cell Transcriptome Analysis Reveals Intratumoral Heterogeneity in ccRCC, which Results in Different Clinical Outcomes》，就是选取ccRCC的3个病例的21个样本（12个肿瘤，9个对照），质控后总计24550个细胞，使用inferCNV算法可以区分成为7786个非恶性，16764个恶性细胞。但其实我是提出来了一个开放式问题共大家讨论，除了这个inferCNV算法，难道就没有其它方法来比较好的区分细胞的恶性与否吗？希望大家集思广益吧！
现在就给大家演示一下，不使用inferCNV算法仍然是可以判断上皮细胞的恶性与否的方法。
关键就是 subset 函数的使用：
rm(list=ls())
options(stringsAsFactors = F)
library(Seurat)
library(ggplot2)
load(file = 'first_sce.Rdata')
load(file = 'phe-of-first-anno.Rdata')
sce=sce.first
table(phe$immune_annotation)
[email protected]=phe
cells.use <- row.names([email protected])[which(phe$immune_annotation=='epi')]
length(cells.use)
sce <-subset(sce, cells=cells.use)
sce
然后重新走单细胞流程：
完全参考单细胞转录组数据分析的流程：
- 祖传的单个10x样本的seurat标准代码
- 祖传的单个10x样本的seurat标准代码（人和鼠需要区别对待）
- seurat标准流程实例之2个10x样本的项目（GSE135927数据集）
# Epithelial cells (n = 5,581) were subsetted and re-clustered into 26 discrete epithelial clusters
sce <- NormalizeData(sce, normalization.method = "LogNormalize",
scale.factor = 1e6)
GetAssay(sce,assay = "RNA")
sce <- FindVariableFeatures(sce,
selection.method = "vst", nfeatures = 2000)
# 步骤 ScaleData 的耗时取决于电脑系统配置（保守估计大于一分钟）
sce <- ScaleData(sce)
sce <- RunPCA(object = sce, pc.genes = VariableFeatures(sce))
DimHeatmap(sce, dims = 1:12, cells = 100, balanced = TRUE)
ElbowPlot(sce)
sce <- FindNeighbors(sce, dims = 1:15)
# Epithelial cells (n = 5,581) were subsetted and re-clustered into 26 discrete epithelial clusters
res.used <- 1
sce <- FindClusters(object = sce, verbose = T, resolution = res.used)
table([email protected]$seurat_clusters)
检查单细胞分群及其来源于的病人信息
很简单的可视化：
set.seed(123)
sce <- RunTSNE(object = sce, dims = 1:15, do.fast = TRUE)
DimPlot(sce,reduction = "tsne",label=T)
DimPlot(sce,reduction = "tsne",label=T, group.by = "patient_id")
table([email protected]$seurat_clusters)
library(gplots)
tab.1=table([email protected]$seurat_clusters,[email protected]$patient_id)
balloonplot(tab.1)
可以很明显的看到，第1,2,7,14,21,23,25 是跨越病人的聚类情况，如下所示：
根据细胞亚群是否跨越病人存在来区分上皮细胞的恶性与否
代码如下：
# 1,2,7,14,21,23,25 是跨越病人的聚类情况
# Cancer cells were identified in 44 of the original 49 tumor biopsy samples
# All cancer cells (n = 3,754) were re-clustered, resulting in 25 unique clusters (Figures S2A and S2B).
# The non-cancer epithelial cell clusters (n = 16) were further annotated into cell subtype
cancer=c(1,2,7,13,14,21,23,25)
[email protected]$cancer <-ifelse([email protected]$seurat_clusters %in% cancer ,'non-cancer','cancer')
# MAke a table
table([email protected]$cancer)
[email protected]
可以看到，我们得到的恶性细胞数量跟作者文章的一样的：
> table([email protected]$cancer)
cancer non-cancer
3732 1712