Differential gene expression Analysis - sample Size

Question

[David_emir]

Hi All,

I am currently having a data from TCGA LUSC samples. I am planning to conduct differential gene expression analysis between normal vs. Tumor samples. My sample size is,Tumor = ~ 200 and Normal < 10. My question is, can we compare 10 Normal samples with 200 Tumor samples? is that statistically correct or do I get correct results? Is that I am thinking wrong?

Secondly, can I pool in Normal samples from other cancer types to increase the sample size for Normal types? in what context TCGA says a sample as normal?

I might be very stupid in asking this, but I want to put my thoughts at rest. Thanks a lot for your assistance, it is highly appreciated.

Sincerely,

Dave,

Answer 1

[Devon Ryan]

Can we compare 10 Normal samples with 200 Tumor samples?

Yes, that's fine.

Can I pool in Normal samples from other cancer types to increase the sample size for Normal types?

I would be very hesitant to do that. The normal samples will need to be from the exact same tissue and as matched in terms of age and other confounders as possible to those originating from the original sample set. Otherwise you're tanking your statistical power.

If possible, get tumour-normal pairs, which will tend to come from the same patient (in my understanding at least).

[pradyumna Jayaram]

Can we compare 3 normal with 300 tumor sample? CESC in TCGA has only 3 normal samples.

[Devon Ryan]

Yes, 3 samples is the absolute minimum for finding group-level differences.

[pradyumna Jayaram]

great! Thank you very much.

[David_emir]

Thanks a lot, I am clear now.