README.md

metadata

license: apache-2.0
tags:
  - biology
  - genomics
size_categories:
  - 1K<n<10K
task_categories:
  - token-classification

eccDNA Copy-Number Level Prediction (Threshold = 10)

This dataset supports the eccDNA copy-number prediction task,
where the model learns to infer relative copy-number levels directly from eccDNA sequences.

Each entry contains:

• sequence: DNA string (A/T/C/G)
• label: copy-number stratum (e.g., 0 = low, 1 = high)

---

📁 Folder Structure

eccdna_copy_number_prediction/
├── data/
│   └── eccdna_copy_number_prediction.csv
├── hf_data/
│   ├── dataset_info.json
│   └── data.arrow
└── README.md

---

Task Description

Predicting eccDNA copy-number levels is biologically meaningful,
as eccDNA abundance reflects amplification intensity and oncogenic potential.

This dataset is derived from eccDNAdb and contains balanced examples
across multiple copy-number thresholds (e.g., >10 and >4.5).

---

Citation

If you use this dataset, please cite:

@inproceedings{liu2025eccdnamamba,
  title={eccDNAMamba: A Pre-Trained Model for Ultra-Long eccDNA Sequence Analysis},
  author={Zhenke Liu and Jien Li and Ziqi Zhang},
  booktitle={ICML 2025 GenBio Workshop},
  year={2025},
  url={https://openreview.net/forum?id=56xKN7KJjy}
}