---
title: CpGenie
canonical_url: "https://www.modelscope.cn/datasets/User040218/CpGenie"
md_url: "https://www.modelscope.cn/datasets/User040218/CpGenie.md"
repository: User040218/CpGenie
last_updated: 2026-11-01
license: other
storage_size: "7.3 GB"
downloads: 8700
stars: 0
---

# CpGenie

> CpGenie - User040218 在 ModelScope 开源的数据集。CpGenie: 1,001 bp CpG-centered sequences and methylation levels from 50 ENCODE RRBS experiments (per-experiment regression).

User040218/CpGenie 是 ModelScope 魔搭社区上的数据集，存储大小 7.3 GB，采用 other 许可。

- **Repository**: User040218/CpGenie
- **License**: other
- **Storage size**: 7.3 GB
- **Downloads**: 8700
- **Stars**: 0
- **Last updated**: 2026-11-01

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

---

# CpGenie

**1,001 bp sequence contexts and methylation levels of CpG sites from 50 ENCODE RRBS experiments** — regression.

CpGenie predicts the DNA methylation level of CpG sites from their sequence context and uses it to score the effects of non-coding variants. As described by the authors, the input is the 1,001 bp sequence centered on the CpG, with the C of the CpG at position 501.

This release contains the 50 ENCODE RRBS experiments used in the paper, one subset per experiment with train/validation/test splits; 47,795,060 rows in total.

Official dataset description: [CpGenie (GitHub)](https://github.com/gifford-lab/CpGenie). See also: [Models (Zenodo 1466073)](https://zenodo.org/records/1466073).

## Subsets (50)

The 50 subsets are named `<cell line>_<ENCODE experiment>` (for example `GM12878_ENCSR000DEY`) and cover cell lines such as MCF-7, LNCaP, GM12878 and HepG2; each has train/validation/test splits.

50 subsets, 47,795,060 rows in total.

## Format

Parquet. Main columns:

| Column | Type | Description |
|---|---|---|
| `sequence` | string | 1,001 bp DNA sequence (the C of the CpG at position 501) |
| `labels` | float32 | Methylation level |
| `coverage` | int64 | Read coverage of the site |
| `chrom` / `pos` / `strand` | string / int64 | CpG coordinate and strand |
| `split` | string | Partition |

## Usage

Example with [DNALLM](https://github.com/zhangtaolab/DNALLM): subset `GM12878_ENCSR000DEY`, regression, `num_labels=1`.

```python
from modelscope.msdatasets import MsDataset
from dnallm import DNADataset

ds = MsDataset.load("User040218/CpGenie", subset_name="GM12878_ENCSR000DEY")
# tokenizer: the tokenizer of your DNALLM model; set max_length to match its tokenization
data = DNADataset(ds, tokenizer=tokenizer, max_length=1003)
```

## Source

If you use this dataset, please cite:

- Zeng H, Gifford DK. Predicting the impact of non-coding variants on DNA methylation. Nucleic Acids Research 2017;45(11):e99. doi:10.1093/nar/gkx177

## License

Declared as `other`: use of the data follows the terms of the original data providers listed in Source.