Does comparison between two groups still make sense if PCA plot suggest no difference?

Question

[wangziwei0010]

Dear Biostar Community,

I want to comapre the green samples with the red ones and get their differentially expressed genes. Before DEG analysis, I draw a PCA plot, which suggest no difference between those two groups.

My Question is:

1. Although I get some DEGs (n = 1400 when FDR<0.05 & |FC| >= 3 ) after DESeq2 pipeline, are those DEGs still make sense since PCA suggests no difference?
2. What conclusion should I make? The two groups are similar (based on PCA plot), or different (based on DESeq2 results)?

Answer 1

[LChart]

You're only looking at 22% of the variance. There may be a small difference in many genes, or a big difference in a handful of genes. It's certainly worth performing the contrast to be sure, IMO.

[Trivas]

To add to this, you should definitely plot different PCs against each other. For example, I have various datasets where there’s very poor separation comparing PC1 to PC2 but very distinct along PC3.

Answer 2

[swbarnes2]

There very well could be a small number of DE's but not enough to be driving PC1 or PC2.

But more than 1000 genes changing three fold? I'd eyeball some, make sure those are good genes that should not be omitted for having few samples with any counts at all.