In the paper by Prodan et al. 2020, their comparison of different ASV pipelines showed that the sets of ASVs found were not same, and they used a 0.002% minimum relative abundance filter to remove most non-consensus ASVs. Are you using an ASV relative abundance filter? If so, what threshold did you choose? Thanks in advance!
I generally don't filter my feature table for most analysis. I basically agree with the authors of that paper when they say
pipeline-specific biases remained after the application of a typical low-abundance filter. Moreover, the filter also removed around 75% of the 4029 consensus ASVs. While abundance-based filters may remove some of the spurious ASVs, they will also remove many true low-abundance biological features. (emphasis mine)
The modern methods are very good at reducing spurious ASVs, as shown by their benchmark of mock communities at the start of the paper, and I want to keep those real, rare features if possible.
Also, you know reviewer three is going to ask how you choose 0.002% as a threshold then ask you to change it...
Graphing is a different story! When making those stacked bar plots, I often merge features based on taxonomy and filter out all but the top taxa, say 5%, just to keep the graph legible.