R/PodiumChange.R

In maSigPro: Significant Gene Expression Profile Differences in Time Course Gene Expression Data

PodiumChange <- function(get, only.sig.iso=FALSE,  comparison=c("any","groups","specific"), group.name="Ctr", time.points=0)
{
  Model <- get$Model
  get2 <- get$get2

  data <- Model$data
  gen <- Model$gen

  edesign <- Model$design$edesign
  repvect = edesign[,2]

  if(only.sig.iso){
    sig.iso2 <- get2$summary
    gen.sig.iso2 <- as.character(gen[rownames(data)%in%sig.iso2])
    NT2 <- get$NumIso.by.gene
    data.clust <- as.matrix( get2$sig.genes$sig.profiles )
    # Removing monoIsoform Genes (there is not podium change)
    genes.2 <- names(NT2[NT2>1])
    data.clust <- data.clust[gen.sig.iso2%in%genes.2,]
    gen.sig.iso22 <- gen.sig.iso2[gen.sig.iso2%in%genes.2]
    # renaming gen.sig.iso22
    gen.sig.iso2 <- gen.sig.iso22
  }
  else{
    data.clust<-as.matrix(data[gen%in%Model$DSG,])
    sig.iso2<-rownames(data.clust)
    gen.sig.iso2 <- as.character(gen[rownames(data)%in%sig.iso2])
    NT2 <- tapply(sig.iso2, gen.sig.iso2, length)
    # Here, there is not any mDSG because in this analysis it is considered only (>1 iso)
  }

  if (length(gen.sig.iso2) == 0 ) print("No selected genes with more than 1 Isoform")
  else{
    #-------------------------------------------------------
    # Major Isoform identification
    #-------------------------------------------------------

    time.M <- tapply(edesign[,1],repvect,mean)
    if(ncol(edesign)==3){
      name3 = colnames(edesign)[3]
      edesign3 = as.data.frame(edesign[,3:ncol(edesign)])
      colnames(edesign3) = name3
    } else edesign3 = edesign[,3:ncol(edesign)]
    groups.M <- apply(edesign3, 2, function(x){tapply(x,repvect,mean)})

    unic <- unique(gen.sig.iso2)
    Mayor=NULL
    LIST = NULL

    for(i in 1:length(unic))
    {
      zz<-data.clust[gen.sig.iso2==unic[i],]
      M <- MayorIso(zz)
      zzM<-zz[M==1,]
      MzzM <- tapply(zzM,repvect,mean)
      zzm <- zz[M!=max(M),]

      if(is.null(nrow(zzm))) ni=1
      else ni=nrow(zzm)

      if(ni==1) Mzzm=tapply(zzm,repvect,mean)
      else Mzzm <- t(apply(zzm, 1, function(x){tapply(x, repvect,mean)}))

      if(ni==1) dif=MzzM - Mzzm
      else dif <- t(apply(Mzzm, 1, function(x){MzzM-x}))
      # Comparison = "any" ----------------------------------------------
      if(comparison=="any"){
        if( any(dif<0) )
          LIST <- c( LIST, unic[i]) }
      # Comparison = "specific" ----------------------------------------------
      else if(comparison=="specific"){
        col <- groups.M[,colnames(groups.M)==group.name]
        if(ni==1) change <- all(dif[col==1 & time.M==time.points]<0)
        else change <- apply(dif[,col==1 & time.M==time.points],1,function(x){all(x<0)})
        if(any(change)) LIST <- c( LIST, unic[i]) }
      # Comparison = group ----------------------------------------------
      else if(comparison=="group"){
        mayors = NULL
        for (k in 3:ncol(edesign))
        {
          mayors = cbind(mayors, MayorIso(zz[,edesign[,k]==1]))
        }
        #When all columns match, substraction with any of them will be 0:
        if (all(mayors-mayors[,1]!=0) )  LIST <- c( LIST, unic[i]) }
    }

    # lists of genes and isoforms:
    gen.L <- gen.sig.iso2 [gen.sig.iso2 %in% LIST]
    data.L <- data.clust[gen.sig.iso2 %in% LIST,]

    output <- list(LIST, data.L,  gen.L, edesign)
    names(output) <- c("L", "data.L", "gen.L","edesign")
    output
  }
}