Unequal sample sizes Deseq2

Question

[Payal]

Hi,

I have two groups with 3 replicates each for differential gene expression analysis.

Control - Sample1 - Sample2 -Sample3

Experiment -Sample1 -Sample2 -Sample3

However, Sample3 in Control has high duplication rate almost double reads than other samples. For differential gene expression in Deseq2, can I exclude the Sample3 in control and compare Controls (Sample1, Sample2) vs Experiment (Sample1, Sample2, Sample3). Will Deseq2 library size normalization take care of unequal sample sizes?

When I am doing DGE with n of 3 I get really low number of significant DGEs (Differentially expressed genes). But when I do n of 2 I get much higher DEGs. Or should I report differential expression results for n of 2 only. Is n of 2 statistically significant?

Thanks, Payal

Answer 1

[ATpoint]

Yes, that is fine. Unequal sample sizes are not unusual. You can check by PCA (plotPCA after transforming data with vst) if the suspicious sample indeed shows evidence for being an outlier. n=2 is of course far from perfect but if the 3rd samples does more harm than good due to being of low quality then yes, exclude it.

[Payal]

Thanks a lot. I did plotPCA with vst and there were clear batch effects for n of 3. After removing the bad sample and performing DEG with n of 2, I again did plotPCA and then I saw clear PCA plots.