I am using DESeq2 to investigate changing microbial taxa after incubation experiments. I am comparing my Preaddition community to 5 post-incubation communities (different nutrient additions & Control) and looking for taxa that significantly change. I have approximately 49,000 OTUs (Operational Taxonomic Units) in the original count table which includes all of the incubations, and I am individually investigating the different incubations. To remove some of these extra OTUs I run:
> microcosmcountTable4 <- microcosmcountTable4[rowSums(microcosmcountTable4)>0,]
I go from 49,000 OTUs to 8500 OTUs meaning about 40,000 OTUs have zero counts in that particular incubation table. If I do not remove those 40k OTUs I end up getting quite different results.
So, my question stems from the command where I am investigating particular nutrient additions from Microcosm 4 count table. I run the command:
> iron4 <- results(dds4, contrast=c("condition","iron","preaddition"))
This investigates just the iron vs. the preaddition (I believe) and does not include the other nutrient treatments. Should I attempt to remove the extra zero count OTUs that likely come about from subsampling within microcosmcountTable4 since I am looking just at the iron vs. Preaddition? Is filtering at the beginning with the other nutrient additions enough? I am am unsure if there is a way to filter at this point.
Thank you!