---
title: Epiphany
canonical_url: "https://www.modelscope.cn/datasets/User040218/Epiphany"
md_url: "https://www.modelscope.cn/datasets/User040218/Epiphany.md"
repository: User040218/Epiphany
last_updated: 2026-11-01
license: cc-by-4.0
storage_size: "1018 MB"
downloads: 5100
stars: 0
---

# Epiphany

> Epiphany - User040218 在 ModelScope 开源的数据集。Epiphany GM12878 epigenomic inputs and Hi-C contact targets; the web preview shows a row index.

User040218/Epiphany 是 ModelScope 魔搭社区上的数据集，存储大小 1018 MB，采用 cc-by-4.0 许可。

- **Repository**: User040218/Epiphany
- **License**: cc-by-4.0
- **Storage size**: 1018 MB
- **Downloads**: 5100
- **Stars**: 0
- **Last updated**: 2026-11-01

Source: https://www.modelscope.cn/datasets/User040218/Epiphany

---

# Epiphany

**GM12878 epigenomic-track inputs and Hi-C contact targets** — regression.

Epiphany predicts cell-type-specific Hi-C contact maps from 1D epigenomic signals using bidirectional LSTM layers, optionally with a generative adversarial network. As described by the authors, any combination of epigenomic tracks of the target cell type can serve as input, and the output is the Hi-C contact map of the same cell type.

This release contains the original GM12878 numeric training tensors (`inputs_embeds`: epigenomic tracks; `labels`: 20,000 Hi-C contact targets) with train/test splits. The previewable subset `index_default` is a row-level index (1,429 rows); use the code below to load the training tensors.

Official dataset description: [Epiphany (GitHub)](https://github.com/arnavmdas/epiphany). See also: [Models (Zenodo 7795868)](https://zenodo.org/records/7795868).

## Splits

Subset: `index_default`

| Split | Rows |
|---|---:|
| test | 207 |
| train | 1,222 |

## Format

Parquet. Main columns:

| Column | Type | Description |
|---|---|---|
| `chromosome` / `hic_bin_start` / `hic_bin_stop` | string / int64 | Window coordinates (index) |
| `source_file` / `source_row` | string / int64 | Native file and row (index) |

## Usage

Example with [DNALLM](https://github.com/zhangtaolab/DNALLM): subset `default`, regression, `num_labels=20000`.

```python
import json
from modelscope import snapshot_download
from dnallm import DNADataset

repo = "User040218/Epiphany"
root = snapshot_download(repo_id=repo, repo_type="dataset", allow_patterns=["source/native-file-map.json"])
files = json.load(open(f"{root}/source/native-file-map.json", encoding="utf-8"))["default"]
root = snapshot_download(repo_id=repo, repo_type="dataset",
                         allow_patterns=[f["path"] for split in files.values() for f in split])
data = DNADataset.load_local_data(
    {split: [f"{root}/{f['path']}" for f in items] for split, items in files.items()},
    seq_col="sequence", label_col="labels")
data.dataset.set_format("torch", columns=["inputs_embeds", "attention_mask", "labels"])
```

Inputs are numeric features in `inputs_embeds`; no tokenization is needed. Use a model that accepts `inputs_embeds`.

The index subsets shown in the web preview are for browsing and locating rows only; the code above downloads and reads the native files listed in `source/native-file-map.json`.

## Notes

- Only GM12878 is included; other cell lines are not.

## Source

If you use this dataset, please cite:

- Yang R, Das A, Gao VR, et al. Epiphany: predicting Hi-C contact maps from 1D epigenomic signals. Genome Biology 2023;24(1):134. doi:10.1186/s13059-023-02934-9

## License

Declared license: CC-BY-4.0. The terms of the original data providers listed in Source also apply.