Question: Study design in DESeq2
1
gravatar for Krzysztof Kolmus
19 months ago by
Poland Warsaw IIMCB
Krzysztof Kolmus10 wrote:

Dear DESeq2 users,

I have a couple of questions regarding study design. They are probably really basic but such analysis is new for me.

Let's suppose I have got some RNASeq data from three independent biological experiments without technical replicates of the same cell line that was cultured at sparse (S) and dense (D) condition. These differentially cultured cells were treated with different siRNA (two types of siRNA per gene) and some inducers. Moreover, I expect some batch effect on my results. I understand that it may sound a bit complicated, thus, I am posting a simplified study design:

name batch condition
1_D_NT 1 D_NT
1_D_siCtrl_1 1 D_siCtrl_1
1_D_siCtrl_2 1 D_siCtrl_2
1_D_siTarget_1 1 D_siTarget_1
1_D_siTarget_2 1 D_siTarget_2
1_D_Inducer 1 D_Inducer
1_S_NT 1 S_NT
1_S_siCtrl_1 1 S_siCtrl_1
1_S_siCtrl_2 1 S_siCtrl_2
1_S_siTarget_1 1 S_siTarget_1
1_S_siTarget_2 1 S_siTarget_2
1_S_Inducer 1 S_Inducer
2_D_NT 2 D_NT
2_D_siCtrl_1 2 D_siCtrl_1
2_D_siCtrl_2 2 D_siCtrl_2
2_D_siTarget_1 2 D_siTarget_1
2_D_siTarget_2 2 D_siTarget_2
2_D_Inducer 2 D_Inducer
2_S_NT 2 S_NT
2_S_siCtrl_1 2 S_siCtrl_1
2_S_siCtrl_2 2 S_siCtrl_2
2_S_siTarget_1 2 S_siTarget_1
2_S_siTarget_2 2 S_siTarget_2
2_S_Inducer 2 S_Inducer
...     

Initially I wrote down the following code:

dds <- DESeqDataSetFromMatrix(countData = cts,
                              colData = coldata,
                              design = ~ batch + condition)

but I am now wondering whether and how I could enhance the possibility of detecting differentially expressed genes between treatments and the same treatment from two confluence states by modifying the colData file?

Do you think something like this might work:

dds <- DESeqDataSetFromMatrix(countData = cts,
                              colData = coldata,
                              design = ~ batch + confluence + condition)

using the following coldata file:

name batch confluence condition
1_D_NT 1 D NT
1_D_siCtrl_1 1 D siCtrl_1
1_D_siCtrl_2 1 D siCtrl_2
1_D_siTarget_1 1 D siTarget_1
1_D_siTarget_2 1 D siTarget_2
1_D_Inducer 1 D Inducer
1_S_NT 1 S NT
1_S_siCtrl_1 1 S siCtrl_1
1_S_siCtrl_2 1 S siCtrl_2
1_S_siTarget_1 1 S siTarget_1
1_S_siTarget_2 1 S siTarget_2
1_S_Inducer 1 S Inducer
2_D_NT 2 D NT
2_D_siCtrl_1 2 D siCtrl_1
2_D_siCtrl_2 2 D siCtrl_2
2_D_siTarget_1 2 D siTarget_1
2_D_siTarget_2 2 D siTarget_2
2_D_Inducer 2 D Inducer
2_S_NT 2 S NT
2_S_siCtrl_1 2 S siCtrl_1
2_S_siCtrl_2 2 S siCtrl_2
2_S_siTarget_1 2 S siTarget_1
2_S_siTarget_2 2 S siTarget_2
2_S_Inducer 2 S Inducer

... 

I am also wondering on constructing a model matrix but I am not sure how it should look like.

Effectively, I would like to run the results function to get the desired comparisons, such as:

results(dds, contrast = list("D.NT", "S.NT")

so batch effect would be already included.

Many thanks for your help and feedback!

All the best,

Krzysztof

ADD COMMENTlink modified 19 months ago • written 19 months ago by Krzysztof Kolmus10
Answer: Study design in DESeq2
0
gravatar for Michael Love
19 months ago by
Michael Love26k
United States
Michael Love26k wrote:

The difference between the first and second design is that, in the second, you assume that the differences between condition (e.g. target1 vs ctrl1) are the same across your confluence variable. If you want to fit different effects you would use your first design.

Otherwise, yes, using ~batch + condition, and then results(dds, contrast=c("condition","B","A")) will control for batch.

ADD COMMENTlink written 19 months ago by Michael Love26k
Answer: Study design in DESeq2
0
gravatar for Krzysztof Kolmus
19 months ago by
Poland Warsaw IIMCB
Krzysztof Kolmus10 wrote:

Thanks a lot!

ADD COMMENTlink written 19 months ago by Krzysztof Kolmus10
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 334 users visited in the last hour