How to create PCA plot and heatmaps using only significant genes in edgeR ?

Question

[Björn]

How to create PCA plot and heatmaps using only significant genes in edgeR ?

I have 10 genes at signficant different levels between comparison groups. How can I create PCA and heatmaps with only those 10 genes.

Suggestions on R code will be appreciated !

[WouterDeCoster]

Actually, this is not appropriate. If you are taking genes which you a priori know that are different between your samples and use those for a PCA then you will of course get a separation between your experimental groups. That's a circular analysis or "double dipping".

[Kirill Tsyganov]

have a look at varistran R package, specifically at

y <- varistran::vst(counts, design=design)
varistran::plot_heatmap(y, n=50)

[Kevin Blighe]

Hey Björn, Have you not searched for how to do this? There is extensive material across the World Wide Web (and on Biostars) about how to do this. Kevin

[Björn]

Thanks Kevin, I used standard edgeR code for PCA but wanted to create PCA for 50 DE genes out of 2000. Use only 50 genes out of 2000 in PCA plot.

[Kevin Blighe]

Okay. You could just filter your logCPM data to include only your genes of interest, and then regenerate the PCA bi-plot. I have also put code here: A: PCA plot from read count matrix from RNA-Seq

[Carambakaracho]

wouldn't it be more or at least equally interesting to see where the data is in the PCA on all data?

[Kevin Blighe]

A good point, Carambakaracho