基因芯片GEO数据分析流程-limma

最近工作关系，需要重现一个文章的基因芯片数据分析，查找差异基因，花了一天时间跑了limma流程，供大家参考。

论文名字为 Identification of inflammatory mediators in patients with Crohn’s disease unresponsive to anti-TNFα therapy, 是三组之间的差异基因比较。想复现这个图：

我图省事，直接GEO数据库中下载了基因表达量的矩阵文件GSE52746_series_matrix.txt.gz, 直接从matrix 开始分析，用的是rstudio，R4.03

#删除环境中所有变量
rm(list=ls())
Sys.setenv(LANGUAGE = "en") #显示英文报错信息
options(stringsAsFactors = FALSE) #禁止chr转成factor

options("repos" = c(CRAN="https://mirrors.tuna.tsinghua.edu.cn/CRAN/"))
options(BioC_mirror="https://mirrors.ustc.edu.cn/bioc/")
dir.create("GSE52746_DEG1")
setwd("/Users/.../Documents/code_and_graph/gut_microbiome_diff_R") # 转换工作目录
list.files()

#安装Biobase and limma,如果已安装，忽略此步，直接导入
if (!requireNamespace("BiocManager", quietly = TRUE))
  +     install.packages("BiocManager") # 安装limma

BiocManager::install("Biobase") # 安装biobase


#导入包
library(Biobase)
library(limma)

#安装GEOquery包，读取matrix
BiocManager::install("GEOquery") # 如果已安装，忽略这一步
library(GEOquery)


#读取基因表达的series_matrix
exprset <- read.table("GSE52746_series_matrix.txt.gz",
                   comment.char = "!",stringsAsFactors=F,header=T,
                   row.names = 1)

dim(exprset) # 查看基因数和样本数
head(exprset)


#导入课题分组的disgn，可以excel编辑导入
design <- read.table("design.csv",header = T, row.names = 1,sep=",")
design

# 指定比较哪些组，建立对比举证
contrast.matrix<-makeContrasts(aCDwithoutTNF-con,
                               aCDwithTNF-con,
                               iCDwithTNF-con,
                               levels = design)
contrast.matrix 

#拟合线性模型lmFit
fit <- lmFit(exprset,design)

#转换为对比模型contrasts.fit
fit2 <- contrasts.fit(fit, contrast.matrix) 

#经验贝叶斯平滑eBayes
fit2 <- eBayes(fit2)

# 提取每一组的对比结果，coef=1是contrast.matrix中第一组，coef=2是第二组，number=inf是提取所有差异基因
deg1_aCDwithoutTNF <- topTable(fit2,adjust='fdr',coef=1,number=Inf) 
deg2_aCDwithTNF <- topTable(fit2,adjust='fdr',coef=2,number=Inf) 
deg3_iCDwithTNF <- topTable(fit2,adjust='fdr',coef=3,number=Inf) 


# 删选显著性 p<0.05, 差异倍数在1.5倍以上的基因
dif1 <- deg1_aCDwithoutTNF[deg1_aCDwithoutTNF[, "P.Value"]<0.05,] # 显著性筛选
head(dif1)
dim(dif1)
summary(dif1) # 产看logFC的范围
nrow(dif1[abs(dif1[,"logFC"])>log2(1.5),]) # 查看差异倍数>1.5倍的基因数目


# 同理筛选第2组和第3组，这儿也可以写成函数，我图省事就直接罗列了
dif2 <- deg2_aCDwithTNF[deg2_aCDwithTNF[, "P.Value"]<0.05,]
dim(dif2)
summary(dif2)
nrow(dif2[abs(dif2[,"logFC"])>log2(1.5),])

# 筛选第3组差异基因
dif3 <- deg3_iCDwithTNF[deg3_iCDwithTNF[, "P.Value"]<0.05,]
dim(dif3)
summary(dif3)
nrow(dif3[abs(dif3[,"logFC"])>log2(1.5),])

# 找出差异基因，画韦恩图

# 提取差异倍数1.5倍以上的基因信息
dif1_gene <- dif1[abs(dif1[,"logFC"])>log2(1.5),]
dif2_gene <- dif2[abs(dif2[,"logFC"])>log2(1.5),]
dif3_gene <- dif3[abs(dif3[,"logFC"])>log2(1.5),]

# 查看各组差异基因数目
dim(dif1_gene)
dim(dif2_gene)
dim(dif3_gene)

# 提取差异基因的名字，就是行名
ge_a <- row.names(dif1_gene)
ge_b <- row.names(dif2_gene)
ge_c <- row.names(dif3_gene)

# 画韦恩图
# 导入包
library(VennDiagram) #没有包的自己安装

# 画图，保存图形文件
venn.diagram(
  x = list(
    A = ge_a,
    B = ge_b,
    C = ge_c
  ),
  filename = "venn figure.tiff", # 保存为tiff格式的文件名
  col = "transparent",
  fill = c("red", "blue", "green"),
  alpha = 0.5,
  label.col = c("darkred", "white", "darkblue", "white",
                "white", "white", "darkgreen"),
  cex = 2.5,
  fontfamily = "serif",
  fontface = "bold",
  cat.default.pos = "text",
  cat.col = c("darkred", "darkblue", "darkgreen"),
  cat.cex = 2.5,
  cat.fontfamily = "serif",
  cat.dist = c(0.06, 0.06, 0.03),
  cat.pos = 0
)

这是最后的图：

差异基因数目差不多。