We've got some raw KEGG ortholog counts from shotgun metagenome sequencing for a somewhat unique (as least as far as I've seen) dataset. There are multiple different study arms, each with a respective treatment group and a control group. These are diet related studies, so the control group in each arm received a certain diet, while the treatment group received the same diet with an additional food of an interest. The control diets were not the same across the different study arms. Sequencing was performed for each sample prior to receiving the intervention (treatment or control), and afterwards. Note that the subjects undergoing the treatment/control interventions are not paired, so a different set of subjects served as the control.
We are primarily interested in finding genes of interest that are differentially expressed within each study arm's treatment group. We want to use the pre-intervention timepoints to control for the initial state of each subject's gene expression, and the control group samples to control for the effect of background diet that was provided. So far we've looked at DESeq2/edgeR for performing this analysis, but have seen both across this forum and the literature that these tools might not be suited for our use case (the same applies to LEFse, ANCOM-BC, etc).
Does anyone have any thoughts on specific pre-existing tools that might be fit for this use case, or would a bespoke analysis be the best route to go down? We're having trouble finding examples of people using this sort of study design, and specifically in the microbiome metagenome space, so I figured I would ask here.
Any input is appreciated - thanks!