> ## 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.

# ToxPredict-v2: Multi-Endpoint Toxicity Prediction

> ToxPredict-v2 screens small molecules for hepatotoxicity, cardiotoxicity, and mutagenicity before synthesis, returning per-endpoint scores and an overall verdict.

ToxPredict-v2 is a 280M parameter model for multi-endpoint toxicity prediction. Given a SMILES string, it returns per-endpoint risk scores across hepatotoxicity, cardiotoxicity, mutagenicity, and acute toxicity, with an overall verdict.

## At a glance

| Field           | Value                                   |
| --------------- | --------------------------------------- |
| Category        | Drug Discovery                          |
| Parameters      | 280M                                    |
| Version         | v2.3                                    |
| Endpoint        | `POST /v1/models/toxpredict-v2/predict` |
| Typical latency | Under 500 ms per compound               |

## Install

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

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

## When to use

* Screen a compound library before wet-lab assays.
* Filter [MolGen-7](/models/molgen-7) candidates before synthesis.
* Prioritize a hit series by predicted safety profile.

## Example

<CodeGroup>
  ```bash cURL theme={null}
  curl -X POST https://api.bunkerbio.com/v1/models/toxpredict-v2/predict \
    -H "Authorization: Bearer $BUNKER_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "smiles": "CC(=O)NC1=CC=C(O)C=C1",
      "name": "Acetaminophen",
      "endpoints": ["hERG", "DILI", "AMES", "LD50"]
    }'
  ```

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

  r = requests.post(
      "https://api.bunkerbio.com/v1/models/toxpredict-v2/predict",
      headers={"Authorization": f"Bearer {os.environ['BUNKER_API_KEY']}"},
      json={
          "smiles": "CC(=O)NC1=CC=C(O)C=C1",
          "name": "Acetaminophen",
          "endpoints": ["hERG", "DILI", "AMES", "LD50"],
      },
  )
  print(r.json())
  ```
</CodeGroup>

```json Response theme={null}
{
  "hERG_block": {"score": 0.03, "risk": "low"},
  "DILI_risk": {"score": 0.61, "risk": "moderate"},
  "AMES_mutagen": {"result": "negative"},
  "LD50_oral_mg_per_kg": 2400,
  "overall": "CAUTION",
  "flags": ["hepatotoxicity"]
}
```

## Parameters

<ParamField path="smiles" type="string" required>
  SMILES representation of the compound.
</ParamField>

<ParamField path="name" type="string">
  Optional human-readable name for the compound. Echoed in the response.
</ParamField>

<ParamField path="endpoints" type="string[]">
  Endpoints to evaluate. Defaults to all supported endpoints. Valid values: `hERG`, `DILI`, `AMES`, `LD50`, `cardiotox`, `nephrotox`.
</ParamField>

## Response fields

<ResponseField name="hERG_block" type="object">
  Cardiotoxicity risk from hERG channel binding. Contains a `score` between 0 and 1 and a categorical `risk` (`low`, `moderate`, `high`).
</ResponseField>

<ResponseField name="DILI_risk" type="object">
  Drug-induced liver injury risk with `score` and `risk` fields.
</ResponseField>

<ResponseField name="AMES_mutagen" type="object">
  Bacterial mutagenicity result (`positive` or `negative`).
</ResponseField>

<ResponseField name="LD50_oral_mg_per_kg" type="number">
  Predicted oral LD50 in milligrams per kilogram.
</ResponseField>

<ResponseField name="overall" type="string">
  Overall verdict: `CLEAR`, `CAUTION`, or `AVOID`.
</ResponseField>

<ResponseField name="flags" type="string[]">
  Human-readable flags for individual concerns raised across endpoints.
</ResponseField>