> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bunkerbio.com/llms.txt
> Use this file to discover all available pages before exploring further.

# BunkerFold-3: Protein Structure Prediction from Sequence

> BunkerFold-3 predicts 3D protein structure from an amino acid sequence with sub-angstrom accuracy on CASP15 targets and full side-chain resolution.

BunkerFold-3 is a 1.2B parameter protein structure prediction model. Given an amino acid sequence, it returns full 3D atomic coordinates with side chains, per-residue confidence scores, and a predicted fold class.

## At a glance

| Field           | Value                                        |
| --------------- | -------------------------------------------- |
| Category        | Protein                                      |
| Parameters      | 1.2B                                         |
| Version         | v3.1                                         |
| Endpoint        | `POST /v1/models/bunkerfold-3/predict`       |
| Typical latency | \~2 seconds for sequences under 150 residues |

## Install

```bash theme={null}
pip install bunker-fold
```

See [Installation](/installation) for full setup, including `uv` and `poetry`.

## When to use

* Predict structure for a novel sequence with no known homolog.
* Resolve side-chain conformations for docking studies.
* Screen designs from a protein engineering pipeline before wet-lab validation.

## Example

<CodeGroup>
  ```bash cURL theme={null}
  curl -X POST https://api.bunkerbio.com/v1/models/bunkerfold-3/predict \
    -H "Authorization: Bearer $BUNKER_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "sequence": "MVLSPADKTNVKAAWGKVGAHAGEYGAEALERMFLSFPTTKTYFPHFDLSH"
    }'
  ```

  ```python Python theme={null}
  import requests, os

  r = requests.post(
      "https://api.bunkerbio.com/v1/models/bunkerfold-3/predict",
      headers={"Authorization": f"Bearer {os.environ['BUNKER_API_KEY']}"},
      json={"sequence": "MVLSPADKTNVKAAWGKVGAHAGEYGAEALERMFLSFPTTKTYFPHFDLSH"},
  )
  print(r.json())
  ```
</CodeGroup>

```json Response theme={null}
{
  "confidence": 0.97,
  "rmsd": 0.42,
  "residues_resolved": "51/51",
  "fold_class": "globin",
  "time_seconds": 1.8,
  "structure_url": "https://cdn.bunkerbio.com/predictions/abc123.pdb"
}
```

## Parameters

<ParamField path="sequence" type="string" required>
  Amino acid sequence in single-letter IUPAC codes. Maximum length 2,048 residues.
</ParamField>

<ParamField path="options.return_side_chains" default="true" type="boolean">
  Include side-chain atoms in the returned structure.
</ParamField>

<ParamField path="options.templates" type="string[]">
  Optional list of PDB IDs to use as templates. Leave empty for template-free prediction.
</ParamField>

## Response fields

<ResponseField name="confidence" type="number">
  Model-wide confidence score between 0 and 1.
</ResponseField>

<ResponseField name="rmsd" type="number">
  Estimated RMSD in angstroms against the predicted best template.
</ResponseField>

<ResponseField name="residues_resolved" type="string">
  Fraction of residues resolved at atomic detail, formatted as `resolved/total`.
</ResponseField>

<ResponseField name="fold_class" type="string">
  Predicted structural class (for example, `globin`, `immunoglobulin`, `tim_barrel`).
</ResponseField>

<ResponseField name="structure_url" type="string">
  Signed URL to a PDB file with the full predicted structure. Valid for 24 hours.
</ResponseField>