countsBoxplots.R

#' Box-plots of (normalized) counts distribution per sample
#'
#' Box-plots of raw and normalized counts distributions per sample to assess the effect of the normalization
#'
#' @name countsBoxplots
#' @param results results produced by DESeq2 or limma with ChIPuana
#' @param conditions factor vector of the condition from which each sample belongs
#' @param method "DESeq2" or "limma"
#' @param col colors of the boxplots (one per biological condition)
#' @param outfile TRUE to export the figure in a png file
#' @return A boxplots of the raw and normalized counts
#' @author Marie-Agnes Dillies, Hugo Varet and Maëlle Daunesse

countsBoxplots <- function(results, conditions, method,
                           col = c("#f3c300", "#875692", "#f38400", "#a1caf1", "#be0032", "#c2b280", "#848482", "#008856", "#e68fac", "#0067a5", "#f99379", "#604e97"), 
                           outfile=FALSE){
  # extract counts and norm counts
  idx <- which(names(results[[1]]) == "baseMean")
  tmp <- results[[1]][,1:(idx-1)]
  nbSamples <- ncol(tmp)/2
  counts <- removeNull(tmp[,1:nbSamples])
  norm.counts <- tmp[,(nbSamples+1):ncol(tmp)]
  norm.count <- removeNull(norm.counts)
  colnames(norm.counts) <- gsub("norm.", "", colnames(norm.counts))
  
  if (outfile) png(filename="countsBoxplots.png", width=2*min(2200, 1800+800*ncol(norm.counts)/10), height=1800, res=300)
  # raw counts
  d <- stack(as.data.frame(log2(as.matrix(counts))))
  d$conditions <- rep(conditions, each=nrow(counts))
  p1 <- ggplot(d) + 
    geom_boxplot(aes(x=.data$ind, y=.data$values+1, fill=.data$conditions), show.legend=FALSE) +
    labs(fill="") +
    scale_fill_manual(values=col) +
    xlab("Samples") +
    ylab("Raw counts (log scale)") +
    ggtitle("Raw counts distribution") +
    theme_light() +
    theme(axis.text.x=element_text(angle=90, hjust=1, vjust=0.5))
  # norm counts
  if (method=="DESeq2") norm.counts <- log2(as.matrix(norm.counts))
  d <- stack(as.data.frame(norm.counts))
  d$conditions <- rep(conditions, each=nrow(norm.counts))
  p2 <- ggplot(d) + 
    geom_boxplot(aes(x=.data$ind, y=.data$values+1, fill=.data$conditions), show.legend=FALSE) +
    labs(fill="") +
    scale_fill_manual(values=col) +
    xlab("Samples") +
    ylab("Normalized counts (log scale)") +
    ggtitle("Normalized counts distribution") +
    theme_light() +
    theme(axis.text.x=element_text(angle=90, hjust=1, vjust=0.5))
  grid.arrange(p1, p2, nrow=1, ncol=2)
  if (outfile) dev.off()
}