In protViz/SRMService: Report Quantitative Mass Spectrometry Data

\newpage

Proteins Quantified in only one condition

Some proteins were quantified only in one condition. In such a case no p-values or fold change can be computed. Nevertheless, proteins with relatively high intensity in one condition but not present in the other condition can have biological relevance. Figure \@ref(fig:plotPresentInOnlyOneCondition) shows how many protein were not quantified in a condition. The Figure \@ref(fig:barplotIntensitiesSingle) visualizes the most intensive proteins not quantified in the other condition.

library(ggplot2)
ggplot(NAinfo , aes(x= name, y=nrProteins)) +
  geom_bar(stat="identity") +
  geom_text(aes(label=nrProteins), vjust=1.6, color="white", size=3.5) +
  theme_minimal()

tmp1 <- results %>%
  select(TopProteinName,grp2$getConditions()$reference , grp2$getConditions()$condition,log2FC) %>%
  dplyr::filter(is.na(log2FC)) %>%
  arrange( desc(!!sym(grp2$getConditions()$reference)) ,desc(!!sym(grp2$getConditions()$condition)) )

plotBaR<-function(tmp2, name){
  if(nrow(tmp2) > 0){
    tmp3 <- tmp2 %>%
      select(TopProteinName,name)
    x <- setNames(tmp3[,2],tmp3[,1])[1:min(20,nrow(tmp3))]
    barplot(x, las = 2, ylab="Normalized intensity", main= name,cex.names=0.7, border="lightblue")
  }
}


tmp2R <- dplyr::filter(tmp1, !is.na(!!sym(grp2$getConditions()$reference)))
tmp2C <- dplyr::filter(tmp1, !is.na(!!sym(grp2$getConditions()$condition)))


if(nrow(tmp2R) >0 & nrow(tmp2C)> 0){
  par(mfrow=c(1,2))
  par(mar=c(10,4,4,1))
  plotBaR(tmp2R , name= grp2$getConditions()$reference)
  plotBaR(tmp2C, name=  grp2$getConditions()$condition)
}else if(nrow(tmp2R) >0 | nrow(tmp2C)> 0){
  par(mar=c(8,4,4,1))
  plotBaR(tmp2R , name= grp2$getConditions()$reference)
  plotBaR(tmp2C, name=  grp2$getConditions()$condition)
}

Furthermore, to integrate those proteins with proteins which do have a fold change and a q.value, we do also provide and a fold change estimate and q.value for those proteins. To emphasize that these values were not obtained employing fitting a statistical model we call the columns in the output file pseudo.log2FC and pseudo.adj.P.Val. Pseudo fold changes are estimated using the mean of the 10% smallest protein averages instead of the absent group average. The missing q-value is substituted by 0. The Volcano plot in Figure \@ref(fig:pseudoVolcanoplot) envisages all the proteins quantified including those identified only in one of the samples (green).

pseudoData <- grp2$getResultTableWithPseudo()

p1 <- quantable::volcano2GB(pseudoData,
                foldchange ="pseudo.log2FC",
                pvalue = "pseudo.P.Value",
                labels = "proteinID",
                ylab="-log10(P Value)",
                log2FCThresh=grp2$qfoldchange,
                pthresh=grp2$qvalue,
                pseudo = pseudoData$P.Value)
p1<-p1 + labs(tag="A") + theme(legend.position="bottom")


p2 <- quantable::volcano2GB(pseudoData,
                foldchange ="pseudo.log2FC",
                pvalue = "pseudo.adj.P.Val",
                labels = "proteinID",
                ylab="-log10(Q Value)",
                log2FCThresh=grp2$qfoldchange,
                pthresh=grp2$qvalue,
                pseudo = pseudoData$adj.P.Val)
p2<-p2 + labs(tag="B") + theme(legend.position="bottom")

if(length( grp2$special )>0)
{
  p1<-quantable::addSpecialProteins(p1,pseudoData,special=  grp2$special,
                     foldchange ="pseudo.log2FC",
                     pvalue = "pseudo.p.mod",
                     labels = "TopProteinName")

  p2<-quantable::addSpecialProteins(p2,pseudoData,special=  grp2$special,
                     foldchange ="pseudo.log2FC",
                     pvalue = "pseudo.q.mod",
                     labels = "TopProteinName")

}

gridExtra::grid.arrange(p1, p2, nrow=1)