Conversation
| for query_name, query in inference_query_set.queries.items(): | ||
| for chain in query.chains: | ||
| if ( | ||
| chain.molecule_type == MoleculeType.PROTEIN |
There was a problem hiding this comment.
Instead of hard-coding the moltypes, could you make them dependent on MSASettings.moltypes (should be in projects.of3_all_atom.config.dataset_config_components), which we use elsewhere to determine which molecule types are expected to have MSAs.
There was a problem hiding this comment.
Oh, moltypes is attribute of MSASettings not MSAComputationSettings.
So I think it's not accessible yet in the DataModule prepare_data level.
Is there any way to cleanly address this?
But I think it's okay to compute & save for both. Later if moltypes don't include RNA or PROTEIN, this field will be ignored. How do you think?
gnikolenyi
left a comment
gnikolenyi
left a comment
There was a problem hiding this comment.
Thanks so much, Minji, nice start, I added some suggestions.
gnikolenyi
left a comment
gnikolenyi
left a comment
There was a problem hiding this comment.
Sorry to spam, I just noticed that there is an issue with one of the unittests, would you mind also taking a look at that?
openfold3/projects/of3_all_atom/config/dataset_config_components.py
Outdated
Show resolved
Hide resolved
|
This is great Minji! Could you please add a test for testing feature generation using the dummy msa file? I think you could do it by extending / copying this test: https://github.com/aqlaboratory/openfold-3/blob/main/openfold3/tests/test_colabfold_msa.py#L225 |
3 participants
Support dummy main MSA consisted of query sequence
Figure: Complex with and without dummy MSA (PDB: 3WBM)