---
title: DeepSEA
canonical_url: "https://www.modelscope.cn/datasets/User040218/DeepSEA"
md_url: "https://www.modelscope.cn/datasets/User040218/DeepSEA.md"
repository: User040218/DeepSEA
last_updated: 2026-11-01
license: cc-by-4.0
storage_size: "4.3 GB"
downloads: 992
stars: 0
---

# DeepSEA

> DeepSEA - User040218 在 ModelScope 开源的数据集。DeepSEA original training data: 1,000 bp DNA sequences with 919 chromatin feature labels (ENCODE, Roadmap).

User040218/DeepSEA 是 ModelScope 魔搭社区上的数据集，存储大小 4.3 GB，采用 cc-by-4.0 许可。

- **Repository**: User040218/DeepSEA
- **License**: cc-by-4.0
- **Storage size**: 4.3 GB
- **Downloads**: 992
- **Stars**: 0
- **Last updated**: 2026-11-01

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

---

# DeepSEA

**1,000 bp DNA sequences with 919 chromatin feature labels from ENCODE and Roadmap** — multilabel classification.

DeepSEA predicts 919 chromatin features (transcription factor binding, DNase I accessibility and histone marks) from DNA sequence to assess non-coding variants. The authors' group re-released the original DeepSEA training, validation and test matrices on Zenodo (Selene manuscript, case 2), together with the list of features that corresponds to the target columns.

This release converts the matrices to 1,000 bp DNA sequences and 919 binary labels: 4,400,000 training, 8,000 validation and 455,024 test examples. Feature names and their order are in `metadata/target_names.txt`.

Official dataset description: [Zenodo 2214970](https://zenodo.org/records/2214970).

## Splits

Subset: `default`

| Split | Rows |
|---|---:|
| train | 4,400,000 |
| validation | 8,000 |
| test | 455,024 |

## Format

Parquet. Main columns:

| Column | Type | Description |
|---|---|---|
| `sequence` | string | 1,000 bp DNA sequence |
| `labels` | list<float32> | 919 binary chromatin feature labels |
| `source_split` / `source_row` | string / int64 | Original matrix and row |

## Usage

Example with [DNALLM](https://github.com/zhangtaolab/DNALLM): subset `default`, multilabel classification, `num_labels=919`.

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

ds = MsDataset.load("User040218/DeepSEA", subset_name="default")
# tokenizer: the tokenizer of your DNALLM model; set max_length to match its tokenization
data = DNADataset(ds, tokenizer=tokenizer, max_length=1002)
```

## Notes

- The `numeric_input` column is empty (null) in this release and can be ignored.

## Source

If you use this dataset, please cite:

- Zhou J, Troyanskaya OG. Predicting effects of noncoding variants with deep learning–based sequence model. Nature Methods 2015;12(10):931-934. doi:10.1038/nmeth.3547

## License

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