I am trying to figure out the best way to model my experiment with DESeq2 to answer two questions.
I have two salinities 35 and 15, where 35 is the control.
Then I have five populations (Wach, Poc, CC, NJ, Mob). I would not consider any of these population to be a control, but rather want to perform pairwise comparison among all possible pairs.
The questions I am trying to answer:
1) How does the hard clam (Mercenaria mercenaria) respond to low salinity (15 ppt) a.k.a 15 ppt vs 35 ppt
2) Do different populations of hard clams respond differently to low salinity. (Wach_15 vs Poc_15, Wach_15 vs CC_15, CC_15 vs Poc_15, etc.)
The issue I am running into is that DESeq2 wants my populations as factors, and when I look at results, only gives 4 population comparison and chooses one to be the control.
How can I make it easier to allow for pairwise comparison among all population pair combinations ?
Do I need to have an interaction?
Do I need to have groups?
ddsHTSeq.adults <- DESeqDataSetFromHTSeqCount(
sampleTable = sample.adults,
directory = directory,
design = ~ abbrev + salinity + abbrev:salinity)
dds <- DESeq(ddsHTSeq.adults)
#abbrev is the abbreviation of the population
> resultsNames(dds)
[1] "Intercept" "abbrev_Mob_vs_CC" "abbrev_NJ_vs_CC" "abbrev_Poc_vs_CC" "abbrev_Wach_vs_CC"
[6] "salinity_15_vs_35" "abbrevMob.salinity15" "abbrevNJ.salinity15" "abbrevPoc.salinity15" "abbrevWach.salinity15"
Check the section on interactions: https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#interactions This here seems to be a good example where a full factorial design (abbrev_salinity combined into a single factor) is much easier to handle and interpret. For getting all pairwise combinations check contrasts section: https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#contrasts