I've been using qiime2 for a few weeks and I can analyze my paired end16s soil data however I'm experiencing some differences with qiime1 results. When I compared the taxonomy assignments from qiime1 (97% OTU clustering with vsearch and RDP classifier, greengenes db) to qiime2 (dada2 denoising + readily available silva classifier from resources webpage) I saw from the taxonomy barplot that qiime1 had assignments to a lot more different OTUs.
I have checked various posts and adjusted my dada2 settings which improved my assignments on qiime2 pipeline quite a lot (since I had lot more reads passing the filters). In the end, I get a similar amount of total features assigned to different taxa however, qiime2 results give a much better level of specificity (more genus/species level assignments) but from qiime1 I get a few hundred more OTUs that have mostly 1 to 10 features assigned to them. Should I treat the OTUs found with a small number of assigned reads from qiime1 pipeline as potentially false positives? Or maybe change something in the classifier part to include more bacteria that have a small number of features assigned to them? This difference is quite important to me because I'm trying to quantify how many different bacteria I have in my test samples to extrapolate how many samples I would need (a ballpark estimation based on beta diversity analyses) to hit a specific number of different bacteria.
Thanks a lot for your time!