Hi,

Apologies, this might sound like a very basic and already asked question. I am new to DESEQ2 and have a naive question regarding the analyses of a 2 x 2 factorial design and result interpretation.

First, I am interested in the main effect and interaction terms. I am comparing two ecotypes (origin) of a fish species that live in two different salinities (high vs. low) and doing a reciprocal transplant rearing (treatment) experiment to test whether there are any gene expression differences (related to embryo development) between the ecotypes and rearing treatment. I am particularly interested in the interaction term because it would help address whether alternate rearing, i.e. rearing embryos in their non-native salinity environment, has an effect on gene expression-related to embryo development. I would like to know if there are any specific responsive genes.

My design matrix is: ~ecotype+rearing+ecotype:rearing.

My resultsNames are: "Intercept" "ecotype_high_vs_low" "rearing_high_vs_low" "ecotypehigh.rearing_high"

Am I interpreting correctly that "ecotypehigh.rearing_high" is the interaction term? When doing such comparison (or contrast), I found no gene that were differentially expressed, which I take to mean that there no genes are significantly affected due to alternate rearing, but I am concerned that I am not interpreting this correctly. Apologies if this is confusing. The vignette was also not clear to me.

Second, I would like to include the developmental stage of these embryos as a covariate in the analyses to account for variation in development among groups. Would I mainly need to extend the design and include the new variable in the design , i.e. ecotype + developmental stage + rearing etc...

Many thanks for reading, and looking forward to hearing from you! I am happy to clarify further if my queries are unclear.

Best, Len

"ecotypehigh.rearing_high" will basically do a 4 group comparison, and tell you if any genes that change going from high to low change differently depending on rearing. So if you find nothing, then you are correct, the alternate rearing doesn't change how genes change in response to high or low.

You can kind of eyeball this yourself if you want; do a comparison of high versus low in rearing 1 only, and then high versus low in rearing 2. Then take the first fold changes, and subtract the second. If you get something close to 0 for every gene, that agrees with what you found with you interaction design.

Throwing another variable at the end of the design, like "stage" or "batch" is a way for the software to model that there is an extra source of variability. So instead of throwing up its hands, and saying "Gee, there's a lot of variability here, I'll do my best" it might be able to say "Oh, there are different subgroups here, maybe it makes sense that the variability is high. I'll take that into account". You might squeeze out a few more differences that were buried under the differences between stages if you include it in your design.