Loop through columns to generate PCA from DESeq2 data

Question

[russell.stewart.j]

I'd like to generate a PCA of my bulk RNAseq data, coloured by each of my variables in the DESeq2 object "vsd". My current code looks like this (to generate a single plot):

pcaData <- plotPCA(vsd, intgroup=c("Age", "BlastRate"), returnData=TRUE)
percentVar <- round(100 * attr(pcaData, "percentVar"))
ggplot(pcaData, aes(PC1, PC2, color=Age, shape=BlastRate)) +
  geom_point(size=3) +
  xlab(paste0("PC1: ",percentVar[1],"% variance")) +
  ylab(paste0("PC2: ",percentVar[2],"% variance")) +
  geom_text(aes(label=name),hjust=-.2, vjust=0) +
  ggtitle("Principal Component Analysis")

Can anyone suggest a method to loop through and swap "Age" with the other variable columns of vsd?

>head(colData(vsd),1)
DataFrame with 1 row and 14 columns
        LibSize LibDiversity PercMapped      Age SpermStatus   SpConc    SpMot Subject.Group PairedSample FertRate
       <factor>     <factor>   <factor> <factor> <character> <factor> <factor>      <factor>     <factor> <factor>
sRNA_1      Low         High       High    42-46         unk      unk      unk     Male-Male            2      Low
       BlastRate RNABatch LibPrepBatch sizeFactor
        <factor> <factor>     <factor>  <numeric>
sRNA_1       Low        1            6   0.929408

Answer 1

[Basti]

If I understand correctly, you might try :

pcaData <- plotPCA(vsd, intgroup=c("Age", "BlastRate"), returnData=TRUE)
percentVar <- round(100 * attr(pcaData, "percentVar"))

for (k in colnames(pcaData)){
ggplot(pcaData, aes_string("PC1", "PC2", color=k, shape=BlastRate)) +
  geom_point(size=3) +
  xlab(paste0("PC1: ",percentVar[1],"% variance")) +
  ylab(paste0("PC2: ",percentVar[2],"% variance")) +
  geom_text(aes(label=name),hjust=-.2, vjust=0) +
  ggtitle("Principal Component Analysis")
}

You may change colnames(pcaData) to a list of variable you want to test.

[russell.stewart.j]

I tried your suggestion and got this error. I'm not sure why I get this error when running your loop but not when I run my code as above. Notably I'm working in Rstudio so each one of these code blocks is a chunk.

Error in aes_string("PC1", "PC2", color = k, shape = BlastRate) : 
object 'BlastRate' not found

[swbarnes2]

Note that intgroup in the first line needs to contain every element of colData that you might want to include. So you may as put them all in; there's no reason not to.

Also note the difference between aes() and aes_string(), how one takes barewords, and one takes quoted strings.

[Basti]

Yes sorry, BlastRate should be quoted "BlastRate" as noticed by swbarnes2