Is it possible (or does it make sense) to display two different gene sets in one PCA-plot?

Question

[jon.brate]

Crossposted from Bioconductor:

Sorry for this naive question, but I am new to principal component analyses and how to display them. I have two different gene sets, coding and non-coding genes, and count data from two different conditions (four replicates of both). I want to compare how the samples cluster in the two gene sets, i.e. do they have the same expression signatures. I made PCAplots of each gene set separately in DESeq2, but I wonder if it makes sense at all to plot them in the same PCA-plot?

Thanks, Jon

Answer 1

[Shicheng Guo]

It is very easy to understand it. In the traditional PCA analysis, each point in PC1-PC2 dimension represent one sample. Since two of your gene set share same samples, you can not repeatly plot them again in one PC1-PC2 figure. In your situation, you can do two PCA plot and compare the difference and the variance distribution of the PCA (Loading trending)

[jon.brate]

That was my feeling to, thanks for the answer!

[Jean-Karim Heriche]

The two plots would most likely not be comparable because PC1/PC2 would be different between the two PCAs. If you want to see how the samples cluster between two conditions and be able to say something about the similarity/difference then you should analyze the two datasets jointly.