Hi all,
We have performed a RNAseq experiment to understand the differential gene expression upon repression of a gene. We have generated a strain in which the gene of interest is placed under doxycycline (DOX) repressible promoter (genotype: tet90). When the DOX is added to medium, gene gets repressed. DOX also has its own effect on the overall gene expression and hence we also have the RNAseq data for WT strain treated with DOX. WT strain grown on YPD medium is used as control. Please refer below for the complete design matrix for details:
| genotype | medium | replicate |
| WT | YPD | 1 |
| WT | YPD | 2 |
| WT | YPD | 3 |
| WT | Dox | 1 |
| WT | Dox | 2 |
| WT | Dox | 3 |
| tet90 | YPD | 1 |
| tet90 | YPD | 2 |
| tet90 | YPD | 3 |
| tet90 | Dox | 1 |
| tet90 | Dox | 2 |
| tet90 | Dox | 3 |
We would like to see the differential expression for gene repression condition vs normal condition (tet90_Dox vs tet90_YPD). However, we would also like to control for the DOX effect in this comparison. I think that this situation is similar to one of the Michael's reply in recent post on comparison (B-control_B) - (A-control_A) C: help DESeq2 model design. I have tried exploring other similar posts on forums but still could not come up with a design formula. What can be the appropriate design formula to address this problem in DESeq2?
Thanks,
Lakhan.
Thank you for your help Michael. Yes, your interpretation is right. I went through the DESeq2 vignette as well as the help page for results function. Example 2 in the results help page is similar with the case we have in our experiment design. As I understand, the interaction term
genotypetet90.mediumDoxwill help us to identify the DEGs under geneX repression (by DOX treatment) in tet90 strain and at the same time control for the DOX effect in WT strain.Thank you again.
Now that I look at the vignette again, I think the controlling variable comes first after the
~so if you wanted to control for medium that would come first as in~medium + genotype + medium:genotypecorrect? In the vignette in very beginning and in interaction section it says thatdesign= ~ batch + conditioncontrols for batch differences and~genotype + conditioncontrols for genotype.The order doesn't matter except that, when
resultsis run without any arguments, it will test on the _last_ variables in the design, so we typically use designs where nuisance variables are in the beginning and the condition of interest in the end.