Hello,
I have metagenome samples from 24 animals. and for each animal, the samples were taken from three different locations (Ileum, Caecum and Colon). I tested the effect of three treatments (8 animals/treatment) on the metagenome level and Now I want to use DESeq2 tool to find the specific genes that are significantly affected by treatment per gut section. For example Effect of treatment, A compared to Treatment B in Ileum.
Therefore, I used DESeq2 with the following design:
dds$group <- factor(paste0(dds$gutsection, dds$treatment))
design(dds) <- ~ group
dds <- DESeq(dds)
So my questions are
1. Is this a correct design?
resultsNames(dds)
results(dds, contrast=c("group", "Ile.TreatmentB", "Ile.TreatmentA"))2. The above design works fine with no error. However, when I subset the data per gut section and run a DESeq2 analysis with this design; (ILE_dataset,~ Treatment), I get different results! Is not that strange? I saw a post from a while ago with a similar question but I am not sure if this is the right explanation also in my case. http://seqanswers.com/forums/showthread.php?t=48354
I'd appreciate any comments and explanation.
Regards
Mahdi
Thanks Michael for a fast reply. So in case that the within variability is not that much different among gut sections, my current design is ok, right?
Now that I see the different results means that within variability is different among my samples, and it would be better to separate the gut sections and run DESeq2 on them. Am I understand it correctly?
Regards
Mahdi
Yes, and you can visualize the differences with PCA. The choice whether to run all together or in separate batches is up to you, and should be guided by these pieces of evidence.