OpenFold3

Overview

Protein, RNA, or DNA monomer and homomer prediction.
Protein-ligand complex prediction with SMILES or CCD ligands.
Prepared JSON runs for heteromers, mixed assemblies, templates, or advanced schema fields.
Batch processing compatible sequence records.

Modes

Mode	Input shape	When to use it
`polymer_structure` Polymer Structure Default	Uses one selected file/source when file parameters are present.	Generate an OpenFold3 query for one protein, RNA, or DNA sequence as a monomer or homomer.
`protein_ligand_complex` Protein-Ligand Complex	Uses one selected file/source when file parameters are present.	Generate an OpenFold3 query for one protein sequence with one small-molecule ligand.
`custom_query_json` Custom Query JSON	Uses one selected file/source when file parameters are present.	Run a prepared OpenFold3 query JSON for mixed complexes, batch jobs, templates, or advanced schemas.
`batch_input` Batch Input	Consumes a folder or output set; useful for batches and pipeline handoffs.	Run OpenFold3 over supported inputs in the selected source root.

Canonical Job Configuration

These are the fields exposed by the default job configuration for openfold3. They are also returned by GET /api/v1/program/params?program=openfold3 and submitted as the params JSON object to POST /api/v1/job/submit.

Parameter	Type	Modes	What it does
`prediction_name` Prediction Name	Text	Polymer Structure, Protein-Ligand Complex	Used as the generated OpenFold3 query name and output folder. Default: openfold3_prediction
`molecule_type` Molecule Type	Text	Polymer Structure	Choose the polymer type for the generated query. Default: Protein; Options: Protein, RNA, DNA
`sequence` Sequence	Sequence	Polymer Structure	Protein, RNA, or DNA sequence. Whitespace is ignored. Required
`chain_ids` Chain IDs	Text	Polymer Structure	One or more chain IDs for the same sequence; multiple IDs create a homomer. Default: A
`protein_sequence` Protein Sequence	Sequence	Protein-Ligand Complex	Protein sequence for the receptor chain. Whitespace is ignored. Required
`protein_chain_ids` Protein Chain IDs	Text	Protein-Ligand Complex	One or more chain IDs for the protein sequence. Default: A
`ligand_type` Ligand Type	Text	Protein-Ligand Complex	Use SMILES for custom small molecules or CCD for an existing component code. Default: SMILES; Options: SMILES, CCD
`ligand` Ligand	Ligand text	Protein-Ligand Complex	Small molecule to include in the generated complex query. Required
`ligand_chain_ids` Ligand Chain IDs	Text	Protein-Ligand Complex	One or more chain IDs for the same ligand. Default: B
`msa_source` MSA Source	Text	Polymer Structure, Protein-Ligand Complex	Single Sequence runs without hosted MSA services. Precomputed MSA attaches a local alignment file for protein/RNA chains. Default: Single Sequence; Options: Single Sequence, Precomputed MSA
`msa_file` MSA File	MSA file	Polymer Structure, Protein-Ligand Complex	Choose a precomputed A3M, Stockholm, or preparsed NPZ alignment file. Required; Files: .a3m, .sto, .npz; Shown when msa_source is Precomputed MSA
`query_json` Query JSON	Input file	Custom Query JSON	Prepared OpenFold3 query JSON using the official queries/chains schema. SubSeq seq JSON from previous pipeline steps is also accepted. Required; Files: .json
`prediction_count` Predictions	Integer	All modes	How many diffusion samples to generate for each query and seed. Default: 1; Range: 1-25

Advanced configuration fields

Parameter	Type	Modes	What it does
`runner_yaml` Runner YAML	Input file	Custom Query JSON	Optional OpenFold3 runner YAML for custom MSA/template settings. Files: .yaml, .yml
`use_templates` Use Templates	Yes/no	Custom Query JSON	Enable only when the custom query JSON and runner settings provide the needed template inputs. Default: false
`model_seed_count` Model Seeds	Integer	All modes	Run multiple model seeds for broader sampling. Total outputs are predictions times seeds. Default: 1; Range: 1-5

Outputs And Metrics

Predicted structures and confidence JSON files.
Top-ranked samples plus per-sample confidence summaries.
pLDDT is local confidence; PAE supports relative-placement inspection.
pTM is overall assembly confidence; ipTM focuses on interfaces.
Clash flags and ranking scores help triage samples.

Common Examples

Protein monomer: Polymer Structure, Protein, chain A, Single Sequence MSA, one prediction and one seed.
Protein homomer: chain IDs A,B,C with a precomputed MSA.
Protein-ligand: protein chain A, ligand CCD ATP as chain B, several predictions and seeds.

Example API params

{
  "mode": "polymer_structure",
  "prediction_name": "target",
  "molecule_type": "Protein",
  "sequence": "MKTAYIAKQRQISFVKSHFSRQDILDLI",
  "chain_ids": "A",
  "prediction_count": 1,
  "model_seed_count": 1
}

Caveats

Single-sequence runs may be less accurate than runs with strong alignments.
Template-enabled runs can leak known structural information in benchmark-style evaluations.
High local confidence does not guarantee a correct interface.

Advanced Submit

Advanced submit is still available for direct program arguments through POST /api/v1/job/submit-advanced. Prefer canonical configuration unless you need exact low-level arguments or are reproducing a known command line.

Advanced submit can run prepared OpenFold3 JSON for heteromers, templates, constraints, or multi-query jobs.
Use canonical configuration for common polymer and protein-ligand cases.

curl -X POST https://subseq.bio/api/v1/job/submit \
  -H "Authorization: Bearer <api_key>" \
  -F program=openfold3 \
  -F 'params={"mode":"polymer_structure","prediction_name":"target","molecule_type":"Protein","sequence":"MKTAYIAKQRQISFVKSHFSRQDILDLI","chain_ids":"A","predictions":1,"model_seeds":1}'