The parameter mmcif_dir in precompute_alignments.py about openfold

[gahdritz]

The downloads currently include no precomputed alignments---those you'll need to generate yourself. We do provide scripts for downloading the required databases and also a simple script that runs the actual alignment (instructions for which are provided in the README). Let me know if you have any further questions.

(We do plan to release a large dataset of AlphaFold training MSAs soon, however. Stay tuned)

[Wenjian-Ma]

The downloads currently include no precomputed alignments---those you'll need to generate yourself. We do provide scripts for downloading the required databases and also a simple script that runs the actual alignment (instructions for which are provided in the README). Let me know if you have any further questions.

(We do plan to release a large dataset of AlphaFold training MSAs soon, however. Stay tuned)

Yeah, I intend to generate alignments by runing "precompute_alignments.py", but I need to input the path "mmcif_dir/" as a parameter in precompute_alignments.py, then I cannot find the path "mmcif_dir", so I don't know which proteins need to be aligned before training the model.

[gahdritz]

input_dir is just a directory containing target files you want to run the alignment tools on. It can contain .cif mmCIF files, FASTA files, or ProteinNet .core files. If you want to use the same data as AlphaFold and have downloaded pdb_mmcif, then the value of input_dir would be <path_to_dir>/pdb_mmcif/mmcif_files.

[Wenjian-Ma]

get it! Thanks for your reply.