R/Block.GUI.R

In ucl-medical-genomics/ChAMP: Chip Analysis Methylation Pipeline for Illumina HumanMethylation450 and EPIC

if(getRversion() >= "3.1.0") utils::globalVariables(c("myBlock","myLoad","myNorm","probe.features.epic","probe.features","Value","ID","Sample","aggregate","x","y","fitted","loess"))

Block.GUI <- function(Block=myBlock,
                      beta=myNorm,
                      pheno=myLoad$pd$Sample_Group,
                      runDMP=TRUE,
                      compare.group=NULL,
                      arraytype="450K")
{
    message("!!! important !!! Since we just upgrated champ.DMP() function, which is now can support multiple phenotypes. Here in Block.GUI() function, if you want to use \"runDMP\" parameter, and your pheno contains more than two groups of phenotypes, you MUST specify compare.group parameter as compare.group=c(\"A\",\"B\") to get DMP value between group A and group B.")


    if(arraytype == "EPIC") data(probe.features.epic) else data(probe.features)
    probe.features <- probe.features[names(Block$allCLID.v),]

    design <- model.matrix( ~ 0 + pheno)
    fit <- lmFit(Block$avbetaCL.m, design)
    fit2 <- eBayes(fit)
    clusterDMP <- topTable(fit2,coef=1,number=nrow(Block$avbetaCL.m),adjust.method="BH")

    B2 <- makeGRangesFromDataFrame(Block$Block[,1:4])
    A2 <- makeGRangesFromDataFrame(data.frame(seqnames=Block$posCL.m[,2],start=Block$posCL.m[,1],end=Block$posCL.m[,1]))
    C <- as.data.frame(findOverlaps(B2,A2))
    clustertmp <- data.frame(Blockindex=paste("Block",C$queryHits,sep="_"),Block$posCL.m[C$subjectHits,2:1])
    clustertmp <- data.frame(clustertmp,clusterDMP[rownames(clustertmp),])

    message("Generation CpG information")
    cpgtmp <- probe.features[names(Block$allCLID.v[which(Block$allCLID.v %in% rownames(clustertmp))]),c(1,2,5,6,7)]

    if(runDMP)
    {
        tmpbeta <- beta
        tmppheno <- pheno
        if(class(pheno)=="numeric") {
            message("  Your pheno parameter is numeric, champ.DMP() function would calculate linear regression for your CpGs.")
        } else {
            message("  You pheno is ",class(pheno)," type.")
            message("    Your pheno information contains following groups. >>")
            sapply(unique(pheno),function(x) message("    <",x,">:",sum(pheno==x)," samples."))
            
            if(length(unique(pheno))==2) {
                message("  Your pheno contains EXACTLY two phenotypes, which is good, compare.group is not needed.")
            } else {
                message("  Your pheno contains more than 2 phenotypes, please use compare.group to specify only two of them here.")
                if(is.null(compare.group)){
                    stop("  compare.group is needed here, please specify compare.group.")
                } else if (sum(compare.group %in% unique(pheno))==2) {
                    message("  Your compare.group is in accord with your pheno, which is good, now we are about to extract information for your compare.group.")
                    tmpbeta <- beta[,which(pheno %in% compare.group)]
                    tmppheno <- pheno[which(pheno %in% compare.group)]
                } else {
                    stop("  Seems your compare.group is not in accord with your pheno, please recheck your pheno and your compare.group.")
                }
            }
        }
        message("Calculating DMP")
        DMP <- champ.DMP(beta=tmpbeta,
                            pheno=tmppheno,
                            adjPVal=1,
                            adjust.method="BH",
                            compare.group=compare.group,
                            arraytype=arraytype)
        cpgDMP <- DMP[[1]]
        cpgtmp <- data.frame(cpgtmp,cpgDMP[rownames(cpgtmp),c(1,3,4,5)])
    }
        
    
    cpgtmp <- data.frame(Clusterindex=Block$allCLID.v[rownames(cpgtmp)],cpgtmp)
    cpgtmp <- data.frame(Blockindex=clustertmp[as.character(cpgtmp$Clusterindex),"Blockindex"],cpgtmp)
    cpgtmp <- cpgtmp[order(cpgtmp$Blockindex,cpgtmp$MAPINFO),]

    innerblockplot <- function(G,cpgidf,block.idx)
    {
        message("<< Generating dmrplot >>")

        for(i in 1:length(G)) G[[i]] <- data.frame(G[[i]],pheno=names(G)[i])
        X <- do.call(rbind,G)
        X <- cbind(X,showtext=paste("cpgID:",X$ID," Sample:",X$Sample,sep=""))
        p <- plot_ly()
        p <- add_trace(p,data=X, x=~pos, y=~Value,type="scatter", mode = "markers", text =~showtext,color=~pheno,marker=list(opacity = 0.6,size = 4))

        message("<< Dots Plotted >>")

        Fit <- lapply(G,function(h) data.frame(pos=unique(h$pos),ID=unique(h$ID),Mean=aggregate(h$Value,by=list(h$pos),mean)[,2]))
        for(i in 1:length(Fit)) Fit[[i]] <- data.frame(Fit[[i]],pheno=paste(names(Fit)[i],"Mean"))
        df <- data.frame(x = unlist(lapply(Fit, "[[", "pos")),
                         y = unlist(lapply(Fit, "[[", "Mean")),
                         ID = unlist(lapply(Fit, "[[", "ID")),
                         cut = unlist(lapply(Fit, "[[", "pheno")))

        p <- add_trace(p,data=df, x =~x, y =~y, text =~ID, color =~cut,type="scatter", mode="lines+markers",line = list(shape = "linear",width = 3,opacity = 0.3),marker=list(size = 1))

        message("<< Mean line Plotted >>")

        Fit2 <- lapply(G,function(h) data.frame(pos=unique(h$pos),ID=unique(h$ID),Fitted=unique(fitted(loess(h$Value ~ h$pos)))))
        for(i in 1:length(Fit2)) Fit2[[i]] <- data.frame(Fit2[[i]],pheno=paste(names(Fit2)[i],"Loess"))
        df2 <- data.frame(x = unlist(lapply(Fit2, "[[", "pos")),
                         y = unlist(lapply(Fit2, "[[", "Fitted")),
                         ID = unlist(lapply(Fit2, "[[", "ID")),
                         cut = unlist(lapply(Fit2, "[[", "pheno")))
        p <- add_trace(p,data=df2, x =~x, y =~y, text =~ID, color =~cut, type="scatter",line =list(shape ="spline",width=3,opacity=0.3,dash = "dash"),marker=list(size = 1))

        message("<< Loess line Plotted >>")

        if(!is.null(cpgidf))
            p <- add_trace(p,data=cpgidf,x=~pos,y=~y,text=~ID,type="scatter",mode="markers",name="CpG Islands",marker=list(opacity = 0.6,size = 10,color="#8b3a62"))
        p <- layout(p, title = paste("Block",block.idx,sep="_"))
    }
    innertestplot <- function()
    {
        plot_ly(x=1:10,y=1:10)
    }

    app <- shinyApp(
        ui = fluidPage(
                tags$head(tags$style("#blockplot{height:40vh !important;}")),

                theme = shinytheme("readable"),
                titlePanel("Block Overview"),
                
		        sidebarLayout(
                    sidebarPanel(
                        br(),
                        width=3,
                        sliderInput("pvaluebin","P value Cutoff:",min = 0,max = 1,step = 0.025,value = 0.05),
                        sliderInput("mincluserbin","Min Number Clusters:",min = 1,max = 350,step = 1,value = 100),
                        actionButton("go", "Submit"),

                        br(),
                        br(),
                        br(),
                        sliderInput("blockbin","Block Index:",min = 1,max = nrow(Block$Block),step = 1,value = 1),
                        actionButton("blocksearch", "Submit")
                                ),#sidebarPanel

                        mainPanel( 
                             tabsetPanel(
                                 tabPanel("Blocktable",
                                          align = "left",
                                          div(dataTableOutput("blocktable"), style = "font-size:75%")
                                         ),
                                 tabPanel("CpGtable",
                                          align = "left",
                                          div(dataTableOutput("cpgtable"), style = "font-size:75%")
                                         ),
                                 tabPanel("Blockplot",
                                          align = "center",
                                          br(),
                                          fluidRow(
                                                   align = "center",
                                                   br(),
                                                   column(width = 12,
                                                          align = "left",
                                                          div(dataTableOutput("clustertable"), style = "font-size:75%")
                                                         ),#column
                                                   column(width = 12,
                                                          align = "center",
                                                          plotlyOutput("blockplot")
                                                         )#column
                                                  )#fluidRow
                                         )
                                    )#tabsetPanel
                                 )#mainPanel
                             )#sidebarLayout
                    ),#ui
    server = function(input, output){

        Block_repalce <- eventReactive(input$blocksearch,
                    {
                        gc()
                        pvalueCutoff <- as.numeric(input$pvaluebin)
                        mincluserCutoff <- as.numeric(input$mincluserbin)
                        block.idx <- as.numeric(input$blockbin)
                        
                        index <- which(Block$posCL.m[,"Chr"] == Block$Block[block.idx,"chr"] & 
                                       Block$posCL.m[,"Pos"] >= Block$Block[block.idx,"start"] & 
                                       Block$posCL.m[,"Pos"] <= Block$Block[block.idx,"end"])

                       clustersinblock <- clustertmp[which(clustertmp$Blockindex==paste("Block",block.idx,sep="_")),] 
                       cpg <- myBlock$allCLID.v[which(myBlock$allCLID.v %in% rownames(clustersinblock))]

                       cpg.pos <- aggregate(probe.features[names(cpg),"MAPINFO"],by=list(cpg),function(x) c(min(x),max(x),max(x)-min(x)))
                       cpg.pos.final <- cpg.pos[,2]
                       rownames(cpg.pos.final) <- cpg.pos[,1]
                       colnames(cpg.pos.final) <- c("Start","End","Width")

                       cpg.gene <- aggregate(probe.features[names(cpg),"gene"],by=list(cpg),function(x) paste(unique(x),collapse=","))
                       cpg.gene.final <- cpg.gene[,2]
                       names(cpg.gene.final) <- cpg.gene[,1]

                       clustersinblock <- data.frame(clustersinblock[,1:3],
                                                     cpg.pos.final[rownames(clustersinblock),],
                                                     Number=table(cpg)[rownames(clustersinblock)],
                                                     gene=cpg.gene.final[rownames(clustersinblock)],
                                                     clustersinblock[4:9])

                        OSindex <- index[which(substr(names(index),1,2)=="OS")]
                        CPGIindex <- index[which(substr(names(index),1,2)=="CP")]
                        SHindex <- index[which(substr(names(index),1,2)=="SH")]

                        Group <- split(as.data.frame(t(Block$avbetaCL.m[OSindex,])),pheno)
                        CPGIpos <- Block$posCL.m[CPGIindex,1]
                        G <- lapply(Group,function(x) t(x))
                        G <- lapply(G,function(h) data.frame(Sample=rep(colnames(h),each=nrow(h)),ID=rep(rownames(h),ncol(h)),pos=rep(Block$posCL.m[OSindex,1],ncol(h)),Value=as.vector(h)))

                        if(length(CPGIindex)>0)
                            cpgidf <- data.frame(ID=rownames(Block$posCL.m)[CPGIindex],pos=Block$posCL.m[CPGIindex,1],y=1)
                        else
                            cpgidf=NULL

                        list(pvalueCutoff=pvalueCutoff,
                             mincluserCutoff=mincluserCutoff,
                             G=G,
                             cpgidf=cpgidf,
                             block.idx=block.idx,
                             clustersinblock=clustersinblock)
                    })
        Cutoff_repalce <- eventReactive(input$go,
                    {
                        gc()
                        pvalueCutoff <- as.numeric(input$pvaluebin)
                        mincluserCutoff <- as.numeric(input$mincluserbin)
                        myblock <- Block$Block[which(Block$Block$p.valueArea <= pvalueCutoff & Block$Block$L >= mincluserCutoff),]
                        mycpgtable <- cpgtmp[which(cpgtmp$Blockindex %in% rownames(myblock)),]

                        list(pvalueCutoff=pvalueCutoff,
                             mincluserCutoff=mincluserCutoff,
                             myblock=myblock,
                             mycpgtable=mycpgtable)
                    })

        output$blocktable <- renderDataTable({
                                            datatable <- Cutoff_repalce()$myblock
                                            datatable <- data.frame(ID=rownames(datatable),datatable)
                                        },options = list(pageLength=20))
        output$cpgtable <- renderDataTable({
                                            datatable <- Cutoff_repalce()$mycpgtable
                                            datatable <- data.frame(ID=rownames(datatable),datatable)
                                        },options = list(pageLength=20))
        output$blockplot <- renderPlotly({
                                            innerblockplot(Block_repalce()$G,Block_repalce()$cpgidf,Block_repalce()$block.idx)
                                        })
        output$clustertable <- renderDataTable({
                                            datatable <- Block_repalce()$clustersinblock
                                            datatable <- data.frame(ID=rownames(datatable),datatable)
                                        },options = list(pageLength=8))

        }
    )
    runApp(app)
}