> ## 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.

# MolGen-7: Generative Small Molecule Design

> MolGen-7 generates novel small molecules optimized for binding affinity, ADMET properties, and synthetic accessibility against a specified biological target.

MolGen-7 is a 3.4B parameter generative model for small molecule design. Given a target and property constraints, it proposes novel candidate molecules with predicted binding affinity, ADMET scores, and synthetic accessibility.

## At a glance

| Field      | Value                                                      |
| ---------- | ---------------------------------------------------------- |
| Category   | Drug Discovery                                             |
| Parameters | 3.4B                                                       |
| Version    | v7.0                                                       |
| Endpoint   | `POST /v1/models/molgen-7/predict`                         |
| Output     | SMILES strings with predicted affinity and property scores |

## Install

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

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

## When to use

* Generate hit candidates against a validated target.
* Optimize a known scaffold for logP, molecular weight, or synthetic accessibility.
* Expand a chemical library with diverse, novel structures.

## Example

<CodeGroup>
  ```bash cURL theme={null}
  curl -X POST https://api.bunkerbio.com/v1/models/molgen-7/predict \
    -H "Authorization: Bearer $BUNKER_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "target": "EGFR T790M",
      "objective": "binding_affinity",
      "constraints": {
        "logP": {"max": 5},
        "mol_weight": {"max": 500},
        "synthetic_steps": {"max": 6}
      },
      "num_candidates": 4
    }'
  ```

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

  r = requests.post(
      "https://api.bunkerbio.com/v1/models/molgen-7/predict",
      headers={"Authorization": f"Bearer {os.environ['BUNKER_API_KEY']}"},
      json={
          "target": "EGFR T790M",
          "objective": "binding_affinity",
          "constraints": {
              "logP": {"max": 5},
              "mol_weight": {"max": 500},
              "synthetic_steps": {"max": 6},
          },
          "num_candidates": 4,
      },
  )
  print(r.json())
  ```
</CodeGroup>

```json Response theme={null}
{
  "candidates_generated": 4,
  "candidates": [
    {
      "id": "BNK-4821",
      "smiles": "CC1=CC(=C...)",
      "affinity_nM": 2.1,
      "synthetic_accessibility": 3.2,
      "logP": 3.8,
      "tox_clear": true
    }
  ]
}
```

## Parameters

<ParamField path="target" type="string" required>
  Target identifier, either a gene symbol with mutation (`EGFR T790M`) or a UniProt ID.
</ParamField>

<ParamField path="objective" type="string" required>
  Optimization objective. One of `binding_affinity`, `selectivity`, `admet`, or `multi`.
</ParamField>

<ParamField path="constraints" type="object">
  Property constraints applied during generation. Each key accepts `min` and `max` bounds.
</ParamField>

<ParamField path="num_candidates" default="4" type="integer">
  Number of candidate molecules to return. Maximum 32.
</ParamField>

## Response fields

<ResponseField name="candidates_generated" type="integer">
  Number of candidates that passed all constraints.
</ResponseField>

<ResponseField name="candidates" type="array">
  Ranked candidate molecules.

  <Expandable title="candidate fields">
    <ResponseField name="id" type="string">
      Bunker Longevity-assigned candidate ID.
    </ResponseField>

    <ResponseField name="smiles" type="string">
      SMILES representation of the molecule.
    </ResponseField>

    <ResponseField name="affinity_nM" type="number">
      Predicted binding affinity in nanomolar.
    </ResponseField>

    <ResponseField name="synthetic_accessibility" type="number">
      SA score from 1 (easy) to 10 (hard) to synthesize.
    </ResponseField>

    <ResponseField name="tox_clear" type="boolean">
      Whether the candidate clears a fast ToxPredict-v2 pre-screen.
    </ResponseField>
  </Expandable>
</ResponseField>

<Tip>
  Pipe candidates through [ToxPredict-v2](/models/toxpredict-v2) for a full toxicity profile before committing to synthesis.
</Tip>