Differential expression DESeq using GLM for timecourse experiment

[jaymiRNA]

Dear Members
I am currently working on NGS data and performed some differential expression analysis using DESeq. At the moment I am trying to find differentially expressed miRNAs using GLM functionality.
My experimental design consists of 4 samples (4stages of developing flower) with 2 biological replicates each. I would like to identify the differentially expressed miRNA during different developmental stages(Stage 1,2,3 and 4). However, I have some doubts regarding my final data interpretation and here is the code I used and the output.

My Dataset:
>Str(SSSA)
Dataframe: 662 obs. of 8 variables:
$ Flower1 : int 1 1 6 0 0 0 6709 0 3 0 ...
$ Flower1.1: int 1 1 1 0 0 0 2261 2 1 2 ...
$ Flower2 : int 0 4 49 0 0 0 7311 0 4 0 ...
$ Flower2.1: int 1 14 69 0 0 0 11211 0 1 0 ...
$ Flower3 : int 0 8 89 1 0 0 1978 0 0 1 ...
$ Flower3.1: int 0 4 72 0 0 0 5392 0 1 1 ...
$ Flower4 : int 0 1 67 0 0 0 4675 0 0 0 ...
$ Flower4.1: int 0 0 21 0 0 0 4629 0 1 5 ...

My code
groups <- c( rep("F1",2), rep("F2",2), rep("F3",2), rep("F4",2) )
cds <- newCountDataSet(SSSA , groups )
cds <- estimateSizeFactors( cds )
cds <- estimateDispersions( cds, "pooled" )
fit0 <- fitNbinomGLMs ( cds, count ~ 1 )
fit1 <- fitNbinomGLMs ( cds, count ~ groups )
pvals <- nbinomGLMTest( fit1, fit0 )
padjGLM <- p.adjust( pvals, method="BH" )
fit1.pval <- cbind(fit1, padjGLM)

My Result
> str(fit1.pval)
'data.frame': 662 obs. of 7 variables:
$ (Intercept): num 1.23 1.16 2.77 -32.46 NA ...
$ groupsF2 : num -3.17 1.07 2.17 -2.1 NA ...
$ groupsF3 : num -32.569 0.859 2.973 30.973 NA ...
$ groupsF4 : num -31.675 -2.008 2.835 -0.875 NA ...
$ deviance : num 0.746 1.895 2.041 0.234 NA ...
$ converged : logi TRUE TRUE TRUE TRUE NA NA ...
$ pval : num 0.28395 0.44052 0.00622 0.86603 NA ...

My doubts.....
1.In my case, is it better to set the conditions as a data frame instead of a vector- “groups <- c( rep("F1",2), rep("F2",2), rep("F3",2), rep("F4",2) )”?
2.In the code, I calculated the object fit0 setting the value 1 after the tilde -“fit0 <- fitNbinomGLMs ( cds, count ~ 1 )” but I am not very clear.. what count ~ 1 stand for/refers to ?

3.My final results( above) shows 7 variables(including the adjusted pvalue) in the data.frame and I am not clear in interpreting this data. In particular , by using ‘groups’ in calling fit 1 i.e. fit1 <- fitNbinomGLMs ( cds, count ~ groups ),I can’t understand if I really compared F1 against all other stages F2,F3,F4, as I intend or If I made any other comparisions.

Could anyone help me on this?

Jay

[jaymiRNA]

nobody knows?

[Simon Anders]

re 1.: If you pass a data frame, you can name you factor. You used "groups". If you pass just a vector, it will always be "condition".

re 2.: You are testing whether the differences between samples from different groups are significantly stronger than the difference between replicates.

Is this what you want to test?