Problems when using clustered msa in alphafold

[mariagviegas]

Hello,
once I obtain the new clustered msa with your code, I am facing problems when trying to get alphafold to use it. I am locally running the one from github: https://github.com/deepmind/alphafold/tree/main
Like they say, I coppy the a3m file I get from you in the output path inside a 'msas' directory and write '--use_precomputed_msas=true' . However, the alphafold job recomputes a new msa from zero.
When I "redo" a job and just coppy the 'msas' folder from the last job, I can see that the msa step is skipped:
Reading MSA from file /mnt/output/.../msas/bfd_uniref_hits.a3m
Unlike the first folder, the msas folder from the alphafold job contains more files:
Reading MSA from file /mnt/output/.../msas/uniref90_hits.sto
Reading MSA from file /mnt/output/.../msas/mgnify_hits.sto

Could you let me know if there is anything I am missing.
Thank you!

Best,
María

[jimhorne]

I found just pointing my input to the directory that contains the .a3m files rather than an absolute path allowed it to predict from all pre-computed MSAs. I didn't add the --use_precomputer_msas=True flag