GLM model design question

Question

[illinois.ks]

Dear,

I have a question related with my design matrix.

My RNA-seq experimental design looks like this. (paired design)

patient gender drug
sample1 patient1 F Pre
sample2 patinet1 F Post
sample3 patient2 M Pre
sample4 patient2 M Post
sample5 patient3 F Pre
sample6 patient3 F Post
sample7 patient4 F Pre
sample8 patient4 F Post
sample9 patient5 F Pre
sample10 patient5 F Post
sample11 patient6 M Pre
sample12 patient6 M Post

What I want to test is this.

1. DEG responding to drug... (regardless of gender)

2. DEG responding to drug on female, DEG responding to drug on male..

   More ultimately, Do genes repond to the drug differently based on gender? (or not really?) In other words, is gender the significant factor to drug response?

3. Which gender is more sensitive to the drug?

In order to test or answer Q1 , my model looks like this.

> design<- model.matrix(~enzyme + enzyme:gender, data=data)
> colnames(design)

[1] "(Intercept)" "enzymepost" "enzymepre:genderM" "enzymepost:genderM"

> lrt1 <- glmLRT(fabry_fit, coef=2) # enzymepost

# returns the genes responding to the drug.. (Q1)

lrt2 <- glmLRT(fabry_fit, coef=4) # enzymepost:genderM

does it returns that genes responding the drug depending on gender differently ?? In this case, we can answer "DEG responding to drug on female DEG responding to drug on male.. " ?? Q2 ...........

Then, what is the meaing of enzymepre:genderF?

I also found that enzymepost:genderM VS enzymepre:genderM returuns relatively similar results!

Could you please somebody to explain the meaning of this two model?

Again, I would like differentially expressed genes under the drug treatment as well as gender. (e.g. up after drug treatment for male, but not for female.. something like this... )

thanks in advance,

[russhh]

Are you sure it's even worth asking the question 'does a gene's response to the drug differ according to gender' when you've only two male subjects?

[Devon Ryan]

Hello illinois.ks!

It appears that your post has been cross-posted to another site: SEQanswers, where Gordon Smyth already replied.

This is typically not recommended as it runs the risk of annoying people in both communities.