I have a microbiome dataset and I am trying to compare different groups of mice (symptomatic animals, asymptomatic animals and a control group).I already performed some analyses with this dataset, but I am observing a great degree of individual variability among the animals.
Thus, I was wondering if I could get better results by comparing only core microbiomes. However, I am not quite sure about what whould be the best strategy to conduct such analyses. Should I use a core microbiome for all animals, or calculate individual cores for each subgroup, merge them and, then , run my comparisons? Should I use the shared_phylotypes.py script for this analysis?
I am currently using QIIME 1.9.1 and the "core microbiomes" are currently being generated by the compute_core_microbiome.py script. I am working with core 80% prevalence in the group. The central microbiome was generated for each of the groups separately (symptomatic animals, asymptomatic animals and a control group) and then concatenated into a single representative OTU Table.
This strategy allowed, in a Beta Diversity test (Microbiome Analyst - NMDS - PERMANOVA), the separation of the three experimental groups at all taxonomic levels analyzed. We took care to validate the separation of groups, that is, to verify that the use of the core did not provide the separation of groups per se and, therefore, three balanced random groups were created, with equal and random sample quantities of the groups classified as symptomatic animals, asymptomatic animals and a control group. The test was performed in triplicate, modifying the samples used in each group at random. With the establishment of these groups, new core files were created and submitted to the Microbiome Analyst, making it possible to verify in all tests that there was no separation of the groups as observed in our original analyzes.
My biggest question and concern is: what is the validity of these comparisons through the core microbiome of the groups and if it will not bring a statistical bias, even ensuring the non-separation by random groups?