RNASeq DEG number changes if I subset the dataset.

Question

[mariano.atm]

Hi, Thank you for reading my question, this is my first post here and it is my very first time doing RNASeq analysis so, any help will be welcome.

I have a RNASeq data set of 6 samples with 3 replicates each. The samples correspond to 2 genotypes (R and S) and 3 times (t0, t12, t24) In my quest to obtain the differential expression genes (DEG). I like to compare the samples if I use the St0 as reference or if I use St0 and Rt0 as reference. In order to get the DEG I use DESeq2 R's Packages.

When I faced the analysis of using Rt0 as reference I start using only the count table with Rt0 Rt12 and Rt24.

I get 3464 DEG in Rt12 vs Rt0

But if I use the complete count table (St0 St12 St24 Rt0 Rt12 Rt24)

and compare the Rt12 vs Rt0 I get only 1038 genes.

In the design for DESeq analysis I use the colData with only the sample specified (St0 St12 St24 Rt0 Rt12 Rt24) and a formula = ~ sample

If anyone can guide me to understand this difference I will be very much pleased.

Answer 1

[Asaf]

DESeq2 uses all the samples to estimate each gene's variance so if you give it six samples or three samples these estimates are different, this is what causing the difference in your results.

[mariano.atm]

Thank you very much Now my question is How do I know which one is the best approach for me?

[Asaf]

More data = better estimates