> ## 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.

# EpiMark: Epigenomic Annotation from Sequence

> EpiMark predicts histone modifications, methylation state, and chromatin accessibility from sequence context alone, with no experimental epigenomic input needed.

EpiMark is a 420M parameter epigenomic annotation model. Given a genomic region and cell type, it predicts histone marks, CpG methylation, and chromatin accessibility from sequence context alone. No ChIP-seq or ATAC-seq data required.

## At a glance

| Field                | Value                                        |
| -------------------- | -------------------------------------------- |
| Category             | Genomics                                     |
| Parameters           | 420M                                         |
| Version              | v1.1                                         |
| Endpoint             | `POST /v1/models/epimark/predict`            |
| Supported cell types | 300+ ENCODE and Roadmap reference cell lines |

## Install

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

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

## When to use

* Annotate regulatory elements in a locus without running assays.
* Compare predicted epigenomic state across cell types.
* Prioritize non-coding variants by predicted regulatory impact.

## Example

<CodeGroup>
  ```bash cURL theme={null}
  curl -X POST https://api.bunkerbio.com/v1/models/epimark/predict \
    -H "Authorization: Bearer $BUNKER_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "region": "chr11:5225000-5250000",
      "cell_type": "K562",
      "marks": ["H3K27ac", "H3K4me3", "ATAC", "CpG_methyl"]
    }'
  ```

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

  r = requests.post(
      "https://api.bunkerbio.com/v1/models/epimark/predict",
      headers={"Authorization": f"Bearer {os.environ['BUNKER_API_KEY']}"},
      json={
          "region": "chr11:5225000-5250000",
          "cell_type": "K562",
          "marks": ["H3K27ac", "H3K4me3", "ATAC", "CpG_methyl"],
      },
  )
  print(r.json())
  ```
</CodeGroup>

```json Response theme={null}
{
  "H3K27ac": {"state": "active_enhancer", "peaks": 3, "p_value": 1e-8},
  "H3K4me3": {"state": "active_promoter"},
  "ATAC": {"state": "open_chromatin"},
  "CpG_methyl": {"level": 0.12, "state": "hypomethylated"},
  "overall_state": "active_regulatory"
}
```

## Parameters

<ParamField path="region" type="string" required>
  Genomic region in `chrN:start-end` format. Maximum window 100 kb.
</ParamField>

<ParamField path="cell_type" type="string" required>
  ENCODE or Roadmap cell type identifier (for example, `K562`, `GM12878`).
</ParamField>

<ParamField path="marks" type="string[]">
  Marks to predict. Defaults to a standard panel. Valid values include `H3K27ac`, `H3K4me3`, `H3K4me1`, `H3K27me3`, `ATAC`, `DNase`, `CpG_methyl`.
</ParamField>

## Response fields

<ResponseField name="H3K27ac" type="object">
  Prediction for the H3K27ac mark with `state`, `peaks`, and `p_value` fields.
</ResponseField>

<ResponseField name="ATAC" type="object">
  Chromatin accessibility prediction. `state` is one of `open_chromatin` or `closed_chromatin`.
</ResponseField>

<ResponseField name="CpG_methyl" type="object">
  CpG methylation prediction with a fractional `level` and categorical `state`.
</ResponseField>

<ResponseField name="overall_state" type="string">
  Aggregated regulatory state for the region: `active_regulatory`, `poised`, `repressed`, or `quiescent`.
</ResponseField>