Guideline for locally generating MSAs

[sangyeon-hits]

Hello, I'd like to reproduce the model's performance while generating MSAs locally.

I've got colabfold_search and mmseqs2 as the search tools, and uniref30_2302 and colabfold_envdb_202108 as DBs following the white paper.

But I don't know the exact commands and workflow for generating .a3m files the same way the authors did.

In particular, as I understand, I need to pair MSAs of different chains using taxonomy. But I'm not sure the current code includes such because I expect that there should be a file containing the taxonomy annotations, which I don't see:

We then assign taxonomy labels to all UniRef sequences using the taxonomy annotation provided by UniProt.

#32 asked a similar question in addition to the confidentiality of the mmseqs server, but that part seems to have been missed out.

[sangyeon-hits]

Related:

1. Featurizer failed... #4 (comment)
2. Featurizer failed... #4 (comment)