Core features / normalization

Hi there,

I am wondering if anyone has an opinion on whether rarefaction should be performed before assessing core-features (features shared across a percentage of samples)? I’m assessing core-features across two major groupings. Each group includes samples from multiple studies (combined in a meta-analysis). I’m concerned that differences in sequencing depths are impacting the resulting core-features within a group and between groups, especially if a more deeply sequenced study has a large sample size compared with other studies within a group. I’m interested in hearing opinions, as I know there are very strong feelings regarding when rarefaction is appropriate. I’m also not opposed to subsampling across studies to even the number of samples included from each study, but similar to the problem with rarefaction, I’m not a fan of losing information if I don’t have to. I look forward to the discussion.