(Edited for clarity and additional details)
Thank you for reaching out. Since the lab handles bugs and feature requests through the bioBakery forum instead of GitHub, issues there are always auto-closed, although you’re welcome (and encouraged) to share them here instead.
HUMAnN intentionally will not read multiple databases of different sources from the same directory, since it instead uses that functionality for other purposes. Specifically, HUMAnN can work with a protein database that’s been split into chunks for computational efficiency (mapping the reads to them serially). Therefore HUMAnN models a protein database as a folder with one or more database chunks in it (derived from the same input set of protein sequences).
So if you mix unrelated chunks in the same folder HUMAnN will produce an error. However, if you want to have two separate DBs, even if each is only represented by a single chunk, they can be used without problems in separate folders.
To avoid confusion in the future for other users, we now have updated the HUMAnN user manual with the same information as well.
I also noticed the same issue. It is not an issue of having multiple incompatible protein databases in the same directory because the same error occurs even when my custom protein DB is moved into a new directory where it is the only file. It appears to simply be an issue with the file naming alone as when I add “201901b” to the end of the file name for my diamond database I do not get the error.
This is a corollary to the point above - HUMAnN software releases are tied to particular database versions and check for compatibility via file names at runtime. For databases that manifest as a folder of files (e.g. the pangenome dataset) we check that all the files in the folder have the expected version string.