Is it a reasonable to delete all unclassified and uncultured ASVs in species level ?

We want to detect more specifics and interpretable ASVs, therefore, we delete all unclassified and uncultured ASVs at the species level. This set of ASVs is the core insight in our study.
However, according to experience, the genus level was more accurate. Is it a reasonable to delete those ASVs which are unclassified and uncultured at the species level?

I prefer to keep this ASVs since by deleting them one is loosing a lot of potentially interesting information.
Usually I delete only unassigned and assigned only to the Bacteria ASVs

1 Like

I agree with you on deleting them is loosing a lot of potentially interesting information. What do you think the biological meaning of assigned or classified ASVs? Why some ASVs can not be assigned or classified?

My guess is that there are several reasons, for example:

  • ASV is most close to the certain family but there is nothing on it at the genus level
  • Regions of rRNA used in analysis have no difference between genuses, so classifier fails to provide higher resolution.

You’re biasing your data if you do this. It depends on your database and your sample type, but is particularly bad if you use an older one.

Databases are classified using available data, but you’re fighting against unculturable organisms (often anaerobic, depending on your environment often interesting) which cannot be easily annotated at the species level; against older and possibly incorrect annotations; against a system which isn’t perfect where new databases are needed given the move toward better sequencing. (But, we have what we have what we have now.)

You also have the general problem that phylogeny ≠ taxonomy and phylogeny/taxonomy may or may not describe functional potential or function in disease.

Ultimately, by doing this, I would argue that you’ve wasted a bunch of time doing sequencing and analysis when you could have gotten almost the same result by culturing. The beauty of culture-free analysis is that you’re able to see things without species/genus annotation on a single nucleotide basis and learn from them! The species level naming answers that age old question Romeo asked Juliet: and yes, a :rose: by another name will smell as sweet and an ASV from another database or with different annotation will be just as important in your community - whether or not it gets fully described to species level.

I’m not sure how unambiguously I can discourage you from doing this. I’m also not sure how unambigiously I can discourage you from using species-level identification in general, but I suspect that is a one-woman adventure in jousting at windmills.