After downloading the raw dbs, there is some setup...
# proteinfold_devt
Thomas Litfin
02/23/2026, 9:21 AMAfter downloading the raw dbs, there is some setup in a sub workflow before they are usable
e
Evangelos Karatzas
02/23/2026, 9:21 AMI think this is where things might go differently from default paths to parsing local db paths
t
Thomas Litfin
02/23/2026, 9:22 AMPaths could go wrong here but I thought Jose did recently test this
e
Evangelos Karatzas
02/23/2026, 9:23 AMIf you already got the faintest idea where the code might be handling them differently, let me know.
j
Jose Espinosa-Carrasco
02/23/2026, 9:23 AMcould you send me a
treee
Evangelos Karatzas
02/23/2026, 9:23 AMon it
Evangelos Karatzas
02/23/2026, 9:27 AM Copy code
├── colabfold_envdb
│ ├── colabfold_envdb_202108_db
│ ├── colabfold_envdb_202108_db_aln
│ ├── colabfold_envdb_202108_db_aln.dbtype
│ ├── colabfold_envdb_202108_db_aln.index
│ ├── colabfold_envdb_202108_db.dbtype
│ ├── colabfold_envdb_202108_db.GPU_READY
│ ├── colabfold_envdb_202108_db_h
│ ├── colabfold_envdb_202108_db_h.dbtype
│ ├── colabfold_envdb_202108_db_h.index
│ ├── colabfold_envdb_202108_db.index
│ ├── colabfold_envdb_202108_db.lookup
│ ├── colabfold_envdb_202108_db_seq
│ ├── colabfold_envdb_202108_db_seq.dbtype
│ ├── colabfold_envdb_202108_db_seq_h
│ ├── colabfold_envdb_202108_db_seq_h.dbtype
│ ├── colabfold_envdb_202108_db_seq_h.index
│ └── colabfold_envdb_202108_db_seq.index
├── params
│ ├── LICENSE
│ ├── params_model_1_multimer_v3.npz
│ ├── params_model_1.npz
│ ├── params_model_2_multimer_v3.npz
│ ├── params_model_2.npz
│ ├── params_model_2_ptm.npz
│ ├── params_model_3_multimer_v3.npz
│ ├── params_model_3.npz
│ ├── params_model_4_multimer_v3.npz
│ ├── params_model_4.npz
│ ├── params_model_5_multimer_v3.npz
│ └── params_model_5.npz
└── uniref30_2302.db
├── uniref30_2302_db
├── uniref30_2302_db_aln
├── uniref30_2302_db_aln.dbtype
├── uniref30_2302_db_aln.index
├── uniref30_2302_db.dbtype
├── uniref30_2302_db.GPU_READY
├── uniref30_2302_db_h
├── uniref30_2302_db_h.dbtype
├── uniref30_2302_db_h.index
├── uniref30_2302_db.index
├── uniref30_2302_db.lookup
├── uniref30_2302_db_mapping
├── uniref30_2302_db_seq
├── uniref30_2302_db_seq.dbtype
├── uniref30_2302_db_seq_h
├── uniref30_2302_db_seq_h.dbtype
├── uniref30_2302_db_seq_h.index
├── uniref30_2302_db_seq.index
└── uniref30_2302_db_taxonomyj
Jose Espinosa-Carrasco
02/23/2026, 9:28 AMthis is how the data is downloaded using the pipeline, isn’t it?
e
Evangelos Karatzas
02/23/2026, 9:28 AMYep
Evangelos Karatzas
02/23/2026, 9:28 AMExactly the same
j
Jose Espinosa-Carrasco
02/23/2026, 9:29 AMthe pipeline expects this instead:
Copy code
colabfold_db/params/alphafold_params_*👍 1
Jose Espinosa-Carrasco
02/23/2026, 9:29 AMthis is the issue
t
Thomas Litfin
02/23/2026, 9:29 AMThe reason for the nesting was to make it fit more cleanly with all the other modes
Thomas Litfin
02/23/2026, 9:30 AMRather than dumping all params in to 1 directory
j
Jose Espinosa-Carrasco
02/23/2026, 9:30 AMalso, this way different set of parameters can be used
👍 1
e
Evangelos Karatzas
02/23/2026, 9:32 AMBut why would the module downloading these works, and then if you provide the same downloaded paths manually does not? Isn't this a bit not intuitive for the user? At least should be made clear at usage.md on what to change after downloading if you want to use the already downloaded files?
t
Thomas Litfin
02/23/2026, 9:32 AMI think the I nconsistency is a bug
j
Jose Espinosa-Carrasco
02/23/2026, 9:33 AMyes, maybe it is something we have to change, because when downloading the process are connected to work but then we changed the structure of the DBs, I have tested this separately, but not downloading and then running the pipeline with the downloaded files
e
Evangelos Karatzas
02/23/2026, 9:34 AMAnd since the database is this huge...we don't want to frustrate the users, right? xD
t
Thomas Litfin
02/23/2026, 9:35 AMI think it is good idea to change the setup process to mimic the expected structure 😁
🙏 1
e
Evangelos Karatzas
02/23/2026, 9:37 AMShall I create an issue?
j
Jose Espinosa-Carrasco
02/23/2026, 9:38 AMYes, please
e
Evangelos Karatzas
02/23/2026, 9:49 AMDone, thanks! https://github.com/nf-core/proteinfold/issues/494
thankyou 1
j
Jose Espinosa-Carrasco
02/23/2026, 11:48 AMWould you mind sending me the content of the work directory of the
COLABFOLD_BATCHJose Espinosa-Carrasco
02/23/2026, 11:49 AMIf you don’t have it handy anymore don’t worry I can try to reproduce it
e
Evangelos Karatzas
02/23/2026, 11:52 AMBecause it was a symlink, it has now been updated to the latest correct version, even in the previous erroneous work folder. But I remember it was just params/alphafold_params_colab_2022-12-06 inside the params folder, but because it didnt exist, it produced the error
👍🏻 1
Evangelos Karatzas
02/23/2026, 11:52 AMThe rest of the folders were as intended
Evangelos Karatzas
02/23/2026, 11:54 AMSo from this:
Copy code
colabfold_alphafold2_params_tags = [
"alphafold2_multimer_v1" : "alphafold_params_colab_2021-10-27",
"alphafold2_multimer_v2" : "alphafold_params_colab_2022-03-02",
"alphafold2_multimer_v3" : "alphafold_params_colab_2022-12-06",
"alphafold2_ptm" : "alphafold_params_2021-07-14"
]Evangelos Karatzas
02/23/2026, 11:54 AMANd now I am not sure if it's even the correct version I renamed xD
Evangelos Karatzas
02/23/2026, 11:55 AMOr if I should have manually downloaded the correct version params for the multimer_v3 mode
2 Views