I have used metastats in mothur to determine which OTUs are differentially abundant between 2 time points in a set of 20 subjects. However, the results include a large number of OTUs that have p values <0.05 but that are very low in abundance or completely absent in some cases. Is there any way to apply some sort of cutoff so that it will only take into account OTUs that are abundant among the subjects. I hope this question makes some sense? Also does the false discovery rate act as an alternative to correction for multiple comparisons, or would I still need to implement this as well?

Thanks a lot for any advice anyone can offer me on this.

The q-values should correct for this. We’d encourage people to check out the original metastats paper for proper interpretation of the output.