写在前面
要帮朋友做RNA-seq的分析，cases vs. controls总共4个样本（2 vs. 2)，看到文献（实验设计比较类似）里用的是EBSeq较为频繁，所以用这个来做。
但从文献来看，EBSeq对样本量依赖较大。
EBSeq
- TPR relatively independent of sample size and presence of outliers.
- Poor FDR control in most situations, relatively unaffected by outliers.
- Medium computational time requirement, increases slightly with sample size.
EBSeq的输入数据是原始的read count，可以通过featureCounts、HTSeq-count等软件包获得。
命令参考链接 http://www.bioconductor.org/packages/release/bioc/html/EBSeq.html
安装
数据不算大，PC应该够用了，所以在window下安装。EBSeq依赖的安装包blockmodeling一直无法成功安装，隔了两天再看，才想到要去仔细看报错文件，发现其中有特殊字符无法识别，索性直接改了，就能直接加载啦。
source("http://bioconductor.org/biocLite.R")
biocLite("EBSeq")
require(EBSeq)
载入需要的程辑包：blockmodeling
Error: package or namespace load failed for ‘blockmodeling’:
attachNamespace()里算'blockmodeling'时.onAttach失败了，详细内容：
调用: parse(con)
错误: 16:28: 意外的INCOMPLETE_STRING
15: journal = "Social Networks",
16: author = as.person("Ale
^
错误: 无法载入程辑包‘blockmodeling’
In addition: Warning message:
In parse(con) :
invalid input found on input connection 'C:/Users/cc/Documents/R/win-library/3.5/blockmodeling/CITATION'
library(blockmodeling)
#出现同样的报错
修改文件'C:/Users/cc/Documents/R/win-library/3.5/blockmodeling/CITATION'的内容，其中对应报错的行有Aleš Žiberna——带帽子的字符，删掉或者改成别的常见字符就OK了。当然如果担心以后出类似问题，改起来琐碎，可以直接设置R的识别语言。
Note:搜索时使用关键词INCOMPLETE_STRING，而非最后的invalid input found on input connection。搜索方向决定了解决问题的效率。
library(EBSeq)
载入需要的程辑包：gplots
载入程辑包：‘gplots’
The following object is masked from ‘package:stats’:
lowess
载入需要的程辑包：testthat
实战
Example
#example
data(GeneMat)
head(GeneMat)
#[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
#Gene_1 1879 2734 2369 2636 2188 9743 9932 10099 9829 9831
#Gene_2 24 40 22 27 31 118 108 144 117 113
Sizes=MedianNorm(GeneMat) #EBSeq requires the library size factor ls for each sample s. Here, ls may be obtained via the function MedianNorm, which reproduces the median normalization approach in DESeq
Conditions=as.factor(rep(c("C1","C2"),each=5)) #设置样本类型
Conditions
#[1] C1 C1 C1 C1 C1 C2 C2 C2 C2 C2
#Levels: C1 C2
EBOut = EBTest(GeneMat,Conditions=as.factor(rep(c("C1","C2"),each=5)),sizeFactors=Sizes, maxround=5)
EBDERes=GetDEResults(EBOut, FDR=0.05)
summary(EBDERes)
str(EBDERes$DEfound) #a list of genes identified with 5% FDR
head(EBDERes$PPMat) # two columns PPEE and PPDE, corresponding to the posterior probabilities of being EE or DE for each gene
str(EBDERes$Status) #EBDERes$Status contains each gene’s status called by EBSeq
实例
paste KD_In1.count KD_In2.count EGFP_In1.count EGFP_In2.count |awk '{print $1"\t"$2"\t"$4"\t"$6"\t"$8}'|sort -k 1 >combined.count
head combined.count
tail combined.count
sed -i '1,2d'combined.count #删除1-2行
sed -i '53800,$d' combined.count #删除53800-last 行
sed -i '1igene\tKD1\tKD2\tEGFP1\tEGFP2' combined.count #增加首行
data=read.table("combined.count",header=T,row.names=1)
head(data)
# EGFP1 EGFP2 KD1 KD2
#ENSMUSG00000000001 281 243 295 233
#ENSMUSG00000000003 0 0 0 0
#ENSMUSG00000000028 20 25 25 12
data=as.matrix(data)
#pre-filtering 这里仅保留了readcounts总和>=10的基因
keep <- rowSums(data[,1:4])>=10
data <- data[keep,]
#跑出结果
require(EBSeq)
Sizes=MedianNorm(data)
EBOut = EBTest(data,Conditions=as.factor(rep(c("EGFP","KD"),each=2)),sizeFactors=Sizes, maxround=15) #maxround迭代次数，一般要求最后结果趋于收敛比较可信
EBDERes=GetDEResults(EBOut, FDR=0.05)
summary(EBDERes)
str(EBDERes$DEfound) #a list of genes identified with 5% FDR
head(EBDERes$PPMat) # two columns PPEE and PPDE, corresponding to the posterior probabilities of being EE or DE for each gene
# PPEE PPDE
#ENSMUSG00000000001 1.0000000 3.203313e-17
#ENSMUSG00000000003 NA NA
str(EBDERes$Status) #EBDERes$Status contains each gene’s status called by EBSeq
#输出DEG(FDR<0.05)
write.table(EBDERes$DEfound,file="DEG_FDR005.txt",quote=F,sep="\t",row.names=F,col.names=F)
#计算 Fold Change (FC) of the raw data as well as the posterior FC of the normalized data.
GeneFC=PostFC(EBOut)
str(GeneFC)
PlotPostVsRawFC(EBOut,GeneFC) #FC vs.Posterior FC的图
write.table(GeneFC,file="FC.txt",quote=F,sep="\t") #注意$Direction
- 如何确定maxround?
查看参数值：EBOut$Alpha,EBOut$P,EBOut$Beta，最后两个值相差<0.01就差不多了。
关于差异表达(RNA-seq)
标准化方法不同+检验不同=多种组合/软件，用之前需要结合自己的样本量来考虑，多参考有相似实验设计的文献，常用的方法都跑一下，自己评估下结果差异，再做定夺。（研究本来就是充满了不确定性，一切都只能用“可能性”来定义，所以，采用同样参数仍然无法完全重复出文献中的结果也是常见。即使你心有芥蒂...）
针对配合meRIP-seq的RNA-seq，因为朋友需要看了几篇文献，基本都是2 vs. 2的样本量，文献里的作法也有差异。
- 直接取了fold change >2 (2017Cancer Cell-m6A Demethylase ALKBH5 Maintains Tumorigenicity of Glioblastoma Stem-like Cells by Sustaining FOXM1 Expression and Cell Proliferation Program)
- EBSeq (2017Cancer Cell-FTO Plays an Oncogenic Role in Acute Myeloid Leukemia as a N6-Methyladenosine RNA Demethylase）
- DESeq2