[dpryan]

Yeah, they just factorized the colData or whatever the equivalent is in the older version that you're using (I'll have to download that version and have a look).

[dpryan]

So the first thing to try is to just reset ecs@designColumns to contain factors (if they don't already). This won't solve the original issue, since I suspect that the conversion to a dataframe is occurring after this step. The quickest way to go about this is to just look at the code for estimatelog2FoldChanges (I'll paste it below) and just execute every line manually. Then you'll know exactly what it's processing and why the duplicate row name error is happening (this will also give you the chance to avert that error then).

estimatelog2FoldChanges <- function(ecs, fitExpToVar="condition", denominator="", getOnlyEffects=FALSE, averageOutExpression=TRUE, nCores=1, quiet=FALSE, file="")
{
   stopifnot(is(ecs, "ExonCountSet"))
   if(any(is.na(sizeFactors(ecs)))){
      stop("Please estimate sizeFactors first\n")}
   if(!fitExpToVar %in% ecs@designColumns){
      stop("fitExpToVar parameter is not in the design columns, double check ecs@designColumns")}
   if(sum(is.na(featureData(ecs)$dispersion))==nrow(counts(ecs))){
      stop("No dispersion parameters found, first call function estimateDispersions...\n")}

   frm <- as.formula(paste("count ~", fitExpToVar,  "* exon"))
   testablegenes <- as.character(unique(fData(ecs)[which(fData(ecs)$testable),]$geneID))

   geteffects <- function(geneID){
     coefficients <- fitAndArrangeCoefs(ecs, geneID=geneID, frm, balanceExons=TRUE)
     if( is.null( coefficients ) ){
        return(coefficients)
     }
     ret <- t(getEffectsForPlotting(coefficients, averageOutExpression=averageOutExpression, groupingVar=fitExpToVar))
     rownames(ret) <- paste(geneID, rownames(ret), sep=":")
     return(ret)
   }

   if( nCores > 1 ){
      if(!is.loaded("mc_fork", PACKAGE="parallel")){
      stop("Please load first parallel package or set parameter nCores to 1...")}
      alleffects <- parallel:::mclapply( testablegenes, function(x){ geteffects(x) }, mc.cores=nCores )
    }else{
      alleffects <- lapply( testablegenes, function(x){geteffects(x)})
    }

    names(alleffects) <- testablegenes
    alleffects <- do.call(rbind, alleffects)
    alleffects <- vst(exp( alleffects ), ecs)
    toadd <- matrix(NA, nrow=nrow(ecs), ncol=ncol(alleffects))
    rownames(toadd) <- featureNames(ecs)

    if( getOnlyEffects ){
       colnames(toadd) <- colnames(alleffects)
       toadd[rownames(alleffects), colnames(alleffects)] <- alleffects
     }else{
       if( denominator == "" ){
          denominator <- as.character(design(ecs, drop=FALSE)[[fitExpToVar]][1])
       }
       stopifnot( any( colnames(alleffects) %in% denominator ) )
       denoCol <- which(colnames(alleffects) == denominator)
       alleffects <- log2(alleffects / alleffects[,denoCol])
       colnames(alleffects) <- sprintf("log2fold(%s/%s)", colnames(alleffects), denominator)
       colnames(toadd) <- colnames(alleffects)
       alleffects <- alleffects[,-denoCol, drop=FALSE]
       toadd <- toadd[,-denoCol, drop=FALSE]
       toadd[rownames(alleffects), colnames(alleffects)] <- alleffects
     }

    fData(ecs) <- cbind(fData(ecs), toadd)
    ecs
}