R/QCinfo.R

In ENmix: Quality control and analysis tools for Illumina DNA methylation BeadChip

QCinfo <- function(rgSet,detPthre=0.000001,detPtype="negative",nbthre=3,samplethre=0.05,CpGthre=0.05,
      bisulthre=NULL,outlier=TRUE,distplot=TRUE)
{

    ##number of bead
    if(!is(rgSet, "rgDataSet") & !is(rgSet, "RGChannelSetExtended"))
      stop("[QCinfo] The input should be an object of 'rgDataSet' or 'RGChannelSetExtended'")

    if(is(rgSet, "rgDataSet")){
      cginfo=getCGinfo(rgSet)
      typeI <- cginfo[cginfo$Infinium_Design_Type %in% c("I","snpI"),]
      typeIred=typeI[typeI$Color_Channel=="Red",]
      typeIgrn=typeI[typeI$Color_Channel=="Grn",]
      typeII<-cginfo[cginfo$Infinium_Design_Type %in% c("II","snpII"),]
      locusNames=c(typeIred$Name,typeIgrn$Name,typeII$Name)
      ##detection P value
      detP<-calcdetP(rgSet,detPtype=detPtype)
      ctrls<-metadata(rgSet)$ictrl
    }else if(is(rgSet, "RGChannelSetExtended")){
      typeI <- getProbeInfo(rgSet, type = "I")
      typeII <- getProbeInfo(rgSet, type = "II")
      locusNames <- getManifestInfo(rgSet, "locusNames")
      ##detection P value
      detP<-detectionP(rgSet)
      ctrls<-getProbeInfo(rgSet,type="Control")
    }

    bc_I <- assays(rgSet)$NBeads[typeI$AddressA,]
    flag<-bc_I>assays(rgSet)$NBeads[typeI$AddressB,]
    bc_I[flag] <- assays(rgSet)$NBeads[typeI$AddressB,][flag];
    nbead <- matrix(NA_real_, ncol = ncol(rgSet), nrow = length(locusNames),
    dimnames = list(locusNames, colnames(rgSet)))
    nbead[typeI$Name,]<-bc_I
    nbead[typeII$Name,]<-assays(rgSet)$NBeads[typeII$AddressA,]
    rm(list=c("bc_I","flag"))

    if(!identical(rownames(detP),rownames(nbead))){
        idx=intersect(rownames(detP),rownames(nbead))
        detP=detP[idx,]
        nbead=nbead[idx,]
    }
    if(!identical(colnames(detP),colnames(nbead))){
        idx=intersect(colnames(detP),colnames(nbead))
        detP=detP[,idx]
        nbead=nbead[,idx]
    }

    ## bisulfite conversion internal controls
    ctrls=ctrls[ctrls$Address %in% rownames(rgSet),]
    ctrl_r <- assays(rgSet)$Red[ctrls$Address,]
    ctrl_g <- assays(rgSet)$Green[ctrls$Address,]
    cc=ctrls[(ctrls$Type %in% c("BISULFITE CONVERSION I")) & ((ctrls$Color %in% 
       c("Green","Lime","LimeGreen")) | (ctrls$ExtendedType %in% 
       c("ctl-BISULFITE-CONVERSION-I-140M_MUS","ctl-BISULFITE-CONVERSION-I-303M_MUS"))),]
    I_green=colMeans(ctrl_g[cc$Address,])
    cc=ctrls[(ctrls$Type %in% c("BISULFITE CONVERSION I")) & ((ctrls$Color %in%
     c("Purple","Red","Tomato")) | (ctrls$ExtendedType %in% 
      c("ctl-BISULFITE-CONVERSION-I-318M_MUS","ctl-BISULFITE-CONVERSION-I-330U_MUS"))),]
    I_red=colMeans(ctrl_r[cc$Address,])
    cc=ctrls[ctrls$Type %in% c("BISULFITE CONVERSION II") & ctrls$Color %in% c("Crimson",
     "DarkMagenta","Red","Orange","Purple","Tomato"),]
    II_red=colMeans(ctrl_r[cc$Address,])
    bisul=(I_green+I_red+II_red)/3

    #threshold of bisulfite conversion control intensity
    if(is.null(bisulthre)){bisulthre=mean(bisul,na.rm=TRUE)-3*sd(bisul,na.rm=TRUE)}

    ##low quality samples
    qcmat <- nbead<nbthre | detP>detPthre
    badValuePerSample <- apply(qcmat,2,sum)/nrow(qcmat)
    flag <- badValuePerSample > samplethre | bisul < bisulthre
    cat(sum(flag)," samples with percentage of low quality CpG value greater than ",
      samplethre, " or bisulfite intensity less than ", bisulthre, "\n")
    badsample=colnames(qcmat)[flag]

    ##low quality CpGs
    qcmat <- qcmat[,!flag]
    NbadValuePerCpG <- apply(qcmat,1,sum)
    badValuePerCpG <- NbadValuePerCpG/ncol(qcmat)
    flag2 <- badValuePerCpG>CpGthre & NbadValuePerCpG>1
    cat(sum(flag2)," CpGs with percentage of low quality value greater than ",
      CpGthre,"\n")
    badCpG <- rownames(qcmat)[flag2]
    qcmat=qcmat[!flag2,]

    #plotting quality scores
    cat("Ploting qc_sample.jpg ...")
    jpeg(filename="qc_sample.jpg",width=1000,height=1000,quality = 100)
    color=rep("black",length(badValuePerSample))
    color[flag]="red"
    plot(badValuePerSample,bisul,xlab="Percent of low quality data per sample",
       ylab="Average bisulfite conversion intensity",cex=1.5,col=color,
       main=paste(length(badsample)," samples were classified as low quality samples"))
    abline(h=bisulthre,lty=2,col="red")
    abline(v=samplethre,lty=2,col="red")
    dev.off()
    cat("Done\n")

    cat("Ploting qc_CpG.jpg ...")
    jpeg(filename="qc_CpG.jpg",width=1000,height=1000,quality = 100)
    par(mfrow=c(2,1))
    hist(badValuePerCpG,breaks=1000,xlab="Percent of low quality data per CpG",
     main=paste(length(badCpG)," CpGs were classified as low quality CpGs"))
    abline(v=CpGthre,lty=2,col="red")
    hist(badValuePerCpG,breaks=1000,xlim=c(0,0.1),
       xlab="Percent of low quality data per CpG",main="Zoom in view")
    abline(v=CpGthre,lty=2,col="red")
    dev.off()
    cat("Done\n")


    if(outlier | distplot){
    if(is(rgSet, "rgDataSet")){mdat=getmeth(rgSet)
    }else if(is(rgSet, "RGChannelSetExtended")){mdat=preprocessRaw(rgSet)}
    rm(rgSet)}

    #Identifying outlier samples
    if(outlier)
    {
    cat("Identifying ourlier samples based on beta or total intensity values...\n")
    mdat=mdat[rownames(qcmat),]
    mdat=mdat[,colnames(qcmat)]
    #outliers based on total intensity values
    mu <- assays(mdat)$Meth+assays(mdat)$Unmeth
    mumean=apply(mu,2,mean,na.rm=TRUE)
    q2575 <- quantile(mumean, probs=c(0.25,0.75), na.rm=TRUE)
    qr <- q2575["75%"]-q2575["25%"]
    low=q2575["25%"] - 3*qr
    flag1=  mumean < low
    cat("After excluding low quality samples and CpGs\n")
    cat(sum(flag1)," samples are outliers based on averaged total intensity value","\n")
rm(mu)
rm(mumean)
    #outliers in beta value distribution
    beta=getB(mdat, type="Illumina")
    qq=apply(beta,2,function(x) quantile(x, probs=c(0.25,0.5,0.75), na.rm=TRUE))
    q2575 <- apply(qq,1,function(x) quantile(x, probs=c(0.25,0.75), na.rm=TRUE))
    qr <- q2575["75%",]-q2575["25%",]
    up=q2575["75%",] + 3*qr
    low=q2575["25%",] - 3*qr
    flag=qq > up | qq < low
    flag=apply(flag,2,sum)>0
    cat(sum(flag)," samples are outliers in beta value distribution","\n")
    flag=flag | flag1
    badsample=c(badsample,colnames(beta)[flag])
    outlier_sample=colnames(beta)[flag]
    cat(sum(flag)," outlier samples were added into badsample list\n")
    if(ncol(qcmat)<15){
    cat("WARNING: Sample size may be too small to correctly identify outlier samples!\n")
    cat("RECOMMAND: set outlier=FALSE or double check total intensity and beta value 
     distribution plots to confirm\n")}
    }

    if(distplot)
    {
    if(!outlier){beta=getB(mdat, type="Illumina")}
    cat("Ploting freqpolygon_beta_beforeQC.jpg ...")
    jpeg(filename="freqpolygon_beta_beforeQC.jpg",width=1000,
     height=500,quality = 100)
    color=rep("black",ncol(beta))
    color[colnames(beta) %in% badsample]="red"
    multifreqpoly(beta,cex.lab=1.4,cex.axis=1.5, col=color, legend="",
    cex.main=1.5,main="Beta value distribution",
    xlab="Methylation beta value")
    dev.off()
    cat("Done\n")

    beta=beta[!(rownames(beta) %in% badCpG),]
    beta=beta[,!(colnames(beta) %in% badsample)]
    cat("Ploting freqpolygon_beta_afterQC.jpg ...")
    jpeg(filename="freqpolygon_beta_afterQC.jpg",width=1000,
     height=500,quality = 100)
    multifreqpoly(beta,cex.lab=1.4,cex.axis=1.5, col="black",legend="",
    cex.main=1.5,main="Beta value distribution",
    xlab="Methylation beta value")
    dev.off()
    cat("Done\n")
    }
rm(mdat)

    if(outlier)
    {list(detP=detP,nbead=nbead,bisul=bisul,badsample=badsample,badCpG=badCpG,
      outlier_sample=outlier_sample)
    }else{list(detP=detP,nbead=nbead,bisul=bisul,badsample=badsample,badCpG=badCpG)}
}