Can I replace the single end sequenced sample with the failed pair end sequenced sample in RNAseq experiment?

Question

[doanphuong19]

Hi all,

I have a RNAseq experiment to measure gene expression with 3 treatment and 3 control, but one of my treatment failed, now I just have 2 treatment vs 3 control. My question is that can I take another treatment sequenced in single end read to replace for one of my failed treatment? I am just beginner and have no experience in this area. I am much appreciated for your help.

[Asaf]

The question is will it add noise or decrease noise. I think that the chance that a gene will be different between the two batches is higher than having a noisy reading in the other two treatments. So yeah, do it 2 vs 3, not too bad, depends on the experiment though.

[doanphuong19]

Thank you very much Asaf, I forgot to mention that my 3 treatments and 3 controls are in pair end reading. Could you please don't mind explain how it depends on the experiment? Thanks a lot.

[Asaf]

What Devon said.

[doanphuong19]

Thanks a lot Asaf, Cheers,

Answer 1

[Devon Ryan]

The problem with adding samples post-hoc is that you're likely going to have a batch effect, due to different preparation dates. Sometimes you get lucky and this isn't a problem, but more often than not you're have a large number of genes different simply due to processing date in the lab. If your biological effect size is very large it may swamp out this effect. On the flip side, if your effect size is moderate or small then I would be very concerned that adding samples would simply increase noise and prevent you from finding anything significant.

Fewer more comparable samples is preferable to more less comparable samples.

[doanphuong19]

Thank you very much Devon, I got what you mean, Cheers