GSVA的使用

• GSVA的简介
  
  Gene Set Variation Analysis，被称为基因集变异分析,是一种非参数的无监督分析方法，主要用来评估芯片核转录组的基因集富集结果。主要是通过将基因在不同样品间的表达量矩阵转化成基因集在样品间的表达量矩阵，从而来评估不同的代谢通路在不同样品间是否富集。其实就是研究这些感兴趣的基因集在不同样品间的差异，或者寻找比较重要的基因集，作为一种分析方法，主要是是为了从生物信息学的角度去解释导致表型差异的原因。它的主要输入文件为表达量的矩阵和基因集的文件，通过gsva的方法就可以得出结果；既可以处理芯片的结果，也可以处理转录组的结果。
• GSVA的安装

source("http://bioconductor.org/biocLite.R")
biocLite('GSVAdata')
biocLite('GSVA')
biocLite("limma")
biocLite("genefilter")

• GSVA的运行注意事项
  • 如果是芯片的数据是需要对数据进行过滤的，可以参考下面的测试数据代码，或者看官方的文档
  • GSVA本身提供了三个算法，一般使用默认的算法就可以了
  • 对于RNA-seq的数据，如果是read count可以选择Possion分布，如果是均一化后的表达量值，可以选择默认参数高斯分布就可以了
  • 读入的数据格式可以参考最后的表达谱数据格式，需要自己制作分组信息
• 模拟数据的使用

#构造一个 30个样本，2万个基因的表达矩阵， 加上 100 个假定的基因集
library(GSVA)

p <- 20000    ## number of genes
n <- 30       ## number of samples
nGS <- 100    ## number of gene sets
min.sz <- 10  ## minimum gene set size
max.sz <- 100 ## maximum gene set size
X <- matrix(rnorm(p*n), nrow=p, dimnames=list(1:p, 1:n))
dim(X)
gs <- as.list(sample(min.sz:max.sz, size=nGS, replace=TRUE)) ## sample gene set sizes
gs <- lapply(gs, function(n, p) sample(1:p, size=n, replace=FALSE), p) ## sample gene sets
es.max <- gsva(X, gs, mx.diff=FALSE, verbose=FALSE, parallel.sz=1)
es.dif <- gsva(X, gs, mx.diff=TRUE, verbose=FALSE, parallel.sz=1)
pheatmap::pheatmap(es.max)
pheatmap::pheatmap(es.dif)

• GSVA的使用
  
  需要如下的包

library(GSEABase)
library(GSVAdata)
data(c2BroadSets)
c2BroadSets

library(Biobase)
library(genefilter)
library(limma)
library(RColorBrewer)
library(GSVA)

#官方文档有数据的预处理过程
cacheDir <- system.file("extdata", package="GSVA")
cachePrefix <- "cache4vignette_"
file.remove(paste(cacheDir, list.files(cacheDir, pattern=cachePrefix), sep="/"))

data(leukemia)
leukemia_eset
head(pData(leukemia_eset))
table(leukemia_eset$subtype)

• 过滤

data(leukemia)
leukemia_eset
filtered_eset <- nsFilter(leukemia_eset, require.entrez=TRUE, remove.dupEntrez=TRUE,var.func=IQR, var.filter=TRUE, var.cutoff=0.5, filterByQuantile=TRUE,feature.exclude="^AFFX")
leukemia_filtered_eset <- filtered_eset$eset

• GSVA的计算差异基因集

cache(leukemia_es <- gsva(leukemia_filtered_eset, c2BroadSets,min.sz=10, max.sz=500, verbose=TRUE),dir=cacheDir, prefix=cachePrefix)

adjPvalueCutoff <- 0.001
logFCcutoff <- log2(2)
design <- model.matrix(~ factor(leukemia_es$subtype))
colnames(design) <- c("ALL", "MLLvsALL")
fit <- lmFit(leukemia_es, design)
fit <- eBayes(fit)
allGeneSets <- topTable(fit, coef="MLLvsALL", number=Inf)
DEgeneSets <- topTable(fit, coef="MLLvsALL", number=Inf,
                       p.value=adjPvalueCutoff, adjust="BH")
res <- decideTests(fit, p.value=adjPvalueCutoff)
summary(res)

• limma计算差异表达基因

logFCcutoff <- log2(2)
design <- model.matrix(~ factor(leukemia_eset$subtype))
colnames(design) <- c("ALL", "MLLvsALL")
fit <- lmFit(leukemia_filtered_eset, design)
fit <- eBayes(fit)
allGenes <- topTable(fit, coef="MLLvsALL", number=Inf)
DEgenes <- topTable(fit, coef="MLLvsALL", number=Inf,
                    p.value=adjPvalueCutoff, adjust="BH", lfc=logFCcutoff)
res <- decideTests(fit, p.value=adjPvalueCutoff, lfc=logFCcutoff)
#summary(res)

• RNAseq的数据
  
  如果是RNA-seq的原始整数的read count 在使用gsva时需要设置kcdf="Possion",如果是取过log的RPKM,TPM等结果可以使用默认的值，所以如果我们在使用fpkm进行分析的时候使用默认参数局可以了

data(commonPickrellHuang)

stopifnot(identical(featureNames(huangArrayRMAnoBatchCommon_eset),featureNames(pickrellCountsArgonneCQNcommon_eset)))

stopifnot(identical(sampleNames(huangArrayRMAnoBatchCommon_eset),sampleNames(pickrellCountsArgonneCQNcommon_eset)))

canonicalC2BroadSets <- c2BroadSets[c(grep("^KEGG", names(c2BroadSets)),grep("^REACTOME", names(c2BroadSets)),grep("^BIOCARTA", names(c2BroadSets)))]

data(genderGenesEntrez)
MSY <- GeneSet(msYgenesEntrez, geneIdType=EntrezIdentifier(),collectionType=BroadCollection(category="c2"), setName="MSY")

XiE <- GeneSet(XiEgenesEntrez, geneIdType=EntrezIdentifier(),collectionType=BroadCollection(category="c2"), setName="XiE")

canonicalC2BroadSets <- GeneSetCollection(c(canonicalC2BroadSets, MSY, XiE))

#使用GSVA方法进行计算
esmicro <- gsva(huangArrayRMAnoBatchCommon_eset, canonicalC2BroadSets, min.sz=5, max.sz=500,mx.diff=TRUE, verbose=FALSE, parallel.sz=1)

esrnaseq <- gsva(pickrellCountsArgonneCQNcommon_eset, canonicalC2BroadSets, min.sz=5, max.sz=500,kcdf="Poisson", mx.diff=TRUE, verbose=FALSE, parallel.sz=1)

• 其实文档中还比较了GSVA和其他不同方法，这里就略过了

实际的表达谱项目数据测试

表达量矩阵格式

gene    AdA3_0h-1    AdA3_0h-2    AdA3_0h-3    AdA3_0h-4    AdA3_0h-5    AdA3_0h-6    AdA3_16h-3    AdA3_16h-4    AdA3_16h-5    AdA3_16h-6    AdGFP_0h-1    AdGFP_0h-2    AdGFP_0h-3    AdGFP_0h-4    AdGFP_0h-5    AdGFP_0h-6    AdGFP_16h-1    AdGFP_16h-2    AdGFP_16h-3    AdGFP_16h-4    AdGFP_16h-5    AdGFP_16h-6
Gnai3    73.842026    52.385326    73.034203    70.679092    67.094292    74.611313    74.853683    72.340401    73.230095    77.39888    70.019714    69.995277    72.172127    72.398514    74.176201    72.700516    60.29203    60.199333    58.043304    58.425671    61.812901    60.219734
Pbsn    0    0    0    0    0    0    0    0    0    0    0    0    0    0    0    0    0    0    0   0
Cdc45    3.653137    5.67695    3.560387    3.287101    4.034892    2.92406    11.427282    14.665288    11.368333    11.96515    4.289696    3.869604    4.055799    3.617629    4.800964    4.790279    20.065876    25.753405    19.732252    26.296944    21.906717    20.378906
Scml2    0.15765    0.022811    0.301977    0.056921    0.022823    0.104651    1.542295    0.370989    0.226674    0.426919    0.132548    0.028549    0.066578    0    0.069393    0.10476    0.981483    1.607109    0.84421    0.912878    1.281105    2.309552

基因集格式

G1/S-Specific Transcription    NA    Rrm2    Ccne1    E2f1    Fbxo5    Cdc6    Dhfr    Cdc45    Pola1    Cdt1    Cdk1    Tyms
E2F mediated regulation of DNA replication    NA    Rrm2    Ccne1    E2f1    Fbxo5    Cdc6    Pola2    Dhfr    Cdc45    Pola1    Cdt1    Cdk1    Ccnb1    Tyms
Formation of Fibrin Clot (Clotting Cascade)    NA    Fga    F11    Fgb    Proc    Serping1    Thbd    Tfpi    F7    Serpine2    Fgg    Serpind1    Serpina5    F8    Pf4
Resolution of D-loop Structures through Holliday Junction Intermediates    NA    Dna2    Bard1    Brca1    Rad51    Rad51c    Brip1    Rad51b    Exo1    Gen1    Wrn    Blm    Brca2    Eme1
Hemostasis    NA    Fga    Serpinf2    Atp1b1    Sparc    Slc16a3    Plaur    Plat    Gpc1    Tgfb1    F11    Fam49b    Arrb2    Fgb    Rapgef4    Proc    Ahsg    Wee1    Vav2    Serping1    Esam    Kif2c    Syk    Thbd    Kif11    Igf1    Tfpi    Rad51c    Serpinb2    Lgals3bp    Dock11    Prkcg    Timp3    Racgap1    Col1a2    Rad51b    Pde5a    Nos3    Dock8    F7    Serpine2    Kif4    Tek    Ehd3    Col1a1    Cav1    Prkch    Fgg    Serpind1    Serpina5    Igf2    Cd74    Ctsw    Irf1    Fcer1g    Pde9a    Plek    Rapgef3    Lcp2    P2rx7    Pde10a    Kif15    Kif23    Csf2rb2    Gnai1    Zfpm1    Slc7a8    Spp2    Kif5a    Il2rg    Lck    Rac2    Hist2h3c1    Gng2    Pla2g4a    Cyb5r1    F8    Gata3    Cd63    Pf4    Serpina3c    Kif3c    P2ry12    Slc16a8    F2rl3    Ppia    Vav3    Ttn    Kif18a    Pde3a    Csf2    Prkcb
Resolution of D-Loop Structures    NA    Dna2    Bard1    Brca1    Rad51    Rad51c    Brip1    Rad51b    Exo1    Gen1    Wrn    Blm    Brca2    Eme1
Common Pathway of Fibrin Clot Formation    NA    Fga    Fgb    Proc    Thbd    Serpine2    Fgg    Serpind1    Serpina5    F8    Pf4
Kinesins    NA    Kif2c    Kif11    Racgap1    Kif4    Kif15    Kif23    Kif5a    Kif3c    Kif18a
Xenobiotics    NA    Cyp2c50    Cyp2c37    Cyp2c54    Cyp2e1    Cyp2c29    Cyp2f2    Cyp2c38    Cyp2b9    Cyp2c55    Arnt2    Cyp2a4
Resolution of D-loop Structures through Synthesis-Dependent Strand Annealing (SDSA)    NA    Dna2    Bard1    Brca1    Rad51    Rad51c    Brip1    Rad51b    Exo1    Wrn    Blm    Brca2
Initial triggering of complement    NA    Crp    C2    C4b    C3    C1qc    C1qa    C1qb    Cfp
Integrin cell surface interactions    NA    Fga    Icam1    Itgb6    Spp1    Fgb    Col5a2    Itga2    Tnc    Col18a1    Col1a2    Itga9    Col1a1    Fgg    Col7a1    Icam2    Fbn1    Itga8    Col8a1    Col4a4
CRMPs in Sema3A signaling    NA    Dpysl3    Dpysl2    Cdk5r1    Dpysl5    Plxna4    Sema3a    Crmp1    Plxna3
Phase 1 - Functionalization of compounds    NA    Adh1    Cyp4a14    Cyp2c50    Cyp2c37    Cyp4a10    Maob    Cyp2c54    Cyp2e1    Cyp4b1    Cyp2c29    Cyp39a1    Cyp2f2    Cyp3a25    Nr1h4    Marc1    Fmo2    Aoc2    Aldh1b1    Cyp2c38    Cmbl    Cyp27b1    Adh7    Aoc3    Cyp2b9    Cyp2c55    Fdx1l    Arnt2    Cyp2a4
Meiotic Recombination    NA    Brca1    Rad51    Hist1h2bc    Hist1h2bg    Blm    Brca2    Hist2h2aa2    Hist2h2aa1    Hist1h2bh    Hist1h2bj    Hist1h2bl    Hist1h2bp    Hist1h2bk    Hist1h2ba    H2afv    Hist3h2a    Hist1h2bn    Hist1h2bf    Hist1h2bm    Hist2h3c1    Hist1h2bb    Prdm9    Msh4    Hist2h2ac    Hist1h4h

具体命令如下

#读取基因集文件
geneSets <- getGmt("test.geneset")
#读取表达量文件并去除重复
mydata <- read.table(file = "all.genes.fpkm.xls",header=T)
name=mydata[,1]
index <- duplicated(mydata[,1])
fildup=mydata[!index,]
exp=fildup[,-1]
row.names(exp)=name
#将数据框转换成矩阵
mydata= as.matrix(exp)

#使用gsva方法进行分析，默认mx.diff=TRUE，min.sz=1,max.zs=Inf，这里是设置最小值和最大值
res_es <- gsva(mydata, geneSets, min.sz=10, max.sz=500, verbose=FALSE, parallel.sz=1)
pheatmap(res_es)

#mx.diff=FALSE es值是一个双峰的分布
es.max <- gsva(mydata, geneSets, mx.diff=FALSE, verbose=FALSE, parallel.sz=1)
pheatmap(es.max)

#mx.diff=TURE es值是一个近似正态分布
es.dif <- gsva(mydata, geneSets, mx.diff=TRUE, verbose=FALSE, parallel.sz=1)
pheatmap(es.dif)

#可以看一下两种不同分布的效果,前者是高斯分布，后者是二项分布
par(mfrow=c(1,2), mar=c(4, 4, 4, 1))
plot(density(as.vector(es.max)), main="Maximum deviation from zero",xlab="GSVA score", lwd=2, las=1, xaxt="n", xlim=c(-0.75, 0.75), cex.axis=0.8)
axis(1, at=seq(-0.75, 0.75, by=0.25), labels=seq(-0.75, 0.75, by=0.25), cex.axis=0.8)
plot(density(as.vector(es.dif)), main="Difference between largest\npositive and negative deviations",xlab="GSVA score", lwd=2, las=1, xaxt="n", xlim=c(-0.75, 0.75), cex.axis=0.8)
axis(1, at=seq(-0.75, 0.75, by=0.25), labels=seq(-0.75, 0.75, by=0.25), cex.axis=0.8)

• limma设置分组的方法

group_list <- factor(c(rep("control",2), rep("siSUZ12",2)))
design <- model.matrix(~group_list)
colnames(design) <- levels(group_list)
rownames(design) <- colnames(counts)

#设置分组
col = names(exp)
sample=col[1:10]
group=c(rep('control',6),rep('treat',4))
phno=data.frame(sample,group)

Group<-factor(phno$group,levels=levels(phno$group))
design<-model.matrix(~0+Group)
colnames(design) <- c("control", "treat")

#获取需要进行差异分析的分组
res=es.max[,1:10]
#定义阈值
logFCcutoff <- log2(1.5)
adjPvalueCutoff <- 0.001
#进行差异分析
fit <- lmFit(res, design)
fit <- eBayes(fit)
allGeneSets <- topTable(fit, coef="treat", number=Inf)
DEgeneSets <- topTable(fit, coef="treat", number=Inf,p.value=adjPvalueCutoff, adjust="BH")
res <- decideTests(fit, p.value=adjPvalueCutoff)
#summary(res)

#画火山图
DEgeneSets$significant="no"
DEgeneSets$significant=ifelse(DEgeneSets$logFC>0|DEgeneSets$logFC<0,"up","down")
ggplot(DEgeneSets,aes(logFC,-1*log10(adj.P.Val)))+geom_point(aes(color =significant)) + xlim(-4,4) + ylim(0,30)+labs(title="Volcanoplot",x="log[2](FC)", y="-log[10](FDR)")+scale_color_manual(values =c("#00ba38","#619cff","#f8766d"))+geom_hline(yintercept=1.3,linetype=4)+geom_vline(xintercept=c(-1,1),linetype=4)

#获取差异基因集的表达量
DEgeneSetspkm = merge(DEgeneSets,es.max,by=0,all.x=TRUE)[,c(1,11:20)]
degsetsp=DEgeneSetspkm[,-1]
name=DEgeneSetspkm[,1]
row.names(degsetsp)=name
pheatmap(degsetsp)