差异表达分析

样本表达量质量评估（剔除离群样本）

图片alt

图片alt

图片alt

提取配对样本

metadata <- count_obj@metadata %>% 
    filter(tissue_type_id %in% c("01", "11")) %>%
    add_count(patient_id, name = "n_patient")%>%
    filter(n_patient == 2) %>%
    arrange(patient_id,tissue_type_id) %>% 
    as.data.frame()
group <- metadata$group %>% as.factor()
paired_counts <- count_obj@count[,metadata$cases]
keep <- rowSums(paired_counts > 0) >= 38
paired_counts <- paired_counts[keep,]

配对样本质量控制

boxplot(log2(paired_counts+1), las = 2, outline = F,  col = group)
limma::plotDensities(log2(paired_counts+1), legend = T,group=group,col = c("green","red"))

图片alt

vst转换配对样本质量控制

library(DESeq2)  
group <- metadata$group %>% as.factor()
colData = data.frame(sample_id = colnames(paired_counts), group = group)
DDS <- DESeq2::DESeqDataSetFromMatrix(paired_counts,
                                      colData = metadata,
                                      design = ~ group)
vst <- DESeq2::vst(DDS)
counts_vst <- assay(vst)

boxplot(counts_vst, las = 2, outline = F, col = group)

limma::plotDensities(counts_vst , legend = F,group=metadata$group)

过滤基因

keep <- rowSums(paired_counts > 0) >= 38

expr <- lnRAN_mRNA(fpkm_obj)@lnRNA[lnRNA_deg_sig$symbol,]
keep <- rowSums(expr > 0.5)>=2 # keep <- rowSums(cpm(y) > 0.5) >= 2
table(keep)

keep <- filterByExpr(exprSet,group = group)

DeSeq2差异表达分析

limma包差异表达分析

makeContrasts(contrasts=cts, levels=design) 1 比较的 -1 被比较的

图片alt

edgeR差异基因分析

# edgeR差异基因分析
library(edgeR)
group <- c(rep(1,length(Norm_sample)),rep(2,length(Tumm_sample)))

y <- DGEList(counts=count_order,group=group)
# 数据过滤
keep <- filterByExpr(y)
y <- y[keep,,keep.lib.size=F]
# 计算标准化因子
y <- calcNormFactors(y)
# 计算离散度
y <- estimateDisp(y)
# 显著性检验
et <- exactTest(y)
# 获取靠前的基因
et <- topTags(et,n=100000)
# 转换为数据框
et <- as.data.frame(et)
# 将行名粘贴为第一列
et <- cbind(rownames(et),et)
str(et)
ggplot(et,aes(x=logFC,y=-log10(FDR)))+
  geom_point()

参考

https://www.jianshu.com/p/e68c95a42cd6
https://www.jianshu.com/p/f009bea514af?utm_campaign=haruki