OpenFold3

How SubSeq Runs It

• Entrypoint is fixed to run_openfold predict.
• The platform forces outputs under /outputs.
• Model assets are managed by SubSeq.
• SubSeq does not expose hosted third-party MSA services in this profile.
• Outputs are written under /outputs/<query_name>/seed_<seed>/; the main model file is usually <query_name>_seed_<seed>_sample_<n>_model.cif.

Required Input

OpenFold3 needs a query JSON file. You can upload it under /inputs or use the UI prefill, which places a starter file at /aux/query.json.

{
  "queries": {
    "ubiquitin": {
      "chains": [
        {
          "molecule_type": "protein",
          "chain_ids": ["A"],
          "sequence": "MQIFVKTLTGKTITLEVEPSDTIENVKAKIQDKEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGG"
        }
      ]
    }
  }
}

Example Arguments

--query-json=/aux/query.json
--use-templates=false
--num-diffusion-samples=1
--num-model-seeds=1

For benchmark-style runs on recently released structures, use --use-templates=false to avoid template leakage.

Notes

• Required local assets are managed by SubSeq.
• Some OpenFold3 CIF outputs omit _atom_site.occupancy; downstream parsers that require it may need a normalization pass that fills occupancy with 1.00.
• This SubSeq image tracks the upstream OpenFold3 CLI with service-specific build adjustments for compatibility.

Submit

Queue a run from New Job -> OpenFold3.