---
title: PrismNet
canonical_url: "https://www.modelscope.cn/datasets/User040218/PrismNet"
md_url: "https://www.modelscope.cn/datasets/User040218/PrismNet.md"
repository: User040218/PrismNet
last_updated: 2026-11-01
license: mit
storage_size: "2.3 GB"
downloads: 51632
stars: 0
---

# PrismNet

> PrismNet - User040218 在 ModelScope 开源的数据集。PrismNet: 172 RBP binding experiments with 101 nt RNA sequences, icSHAPE structure and binding labels.

User040218/PrismNet 是 ModelScope 魔搭社区上的数据集，存储大小 2.3 GB，采用 mit 许可。

- **Repository**: User040218/PrismNet
- **License**: mit
- **Storage size**: 2.3 GB
- **Downloads**: 51632
- **Stars**: 0
- **Last updated**: 2026-11-01

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

---

# PrismNet

**172 RBP binding experiments with 101 nt RNA sequences, icSHAPE structure scores and binding labels** — binary classification.

PrismNet predicts dynamic cellular protein–RNA interactions by combining in vivo RNA structure (icSHAPE) with sequence. The authors' GitHub repository provides data for 172 samples (RBP × cell line) for training and testing, and their website visualizes the icSHAPE data and results.

This release merges the 172 experiments into one subset, `experiments`, with 2,370,394 rows and each experiment's original train/validation/test partition. `experiment` identifies the experiment, `sequence` is the 101 nt sequence, `icshape` the per-base structure score and `labels` the binding label.

Official dataset description: [PrismNet (GitHub)](https://github.com/kuixu/PrismNet). See also: [PrismNet website](http://prismnet.zhanglab.net/).

## Splits

Subset: `experiments`

| Split | Rows |
|---|---:|
| test | 474,026 |
| train | 1,685,570 |
| validation | 210,798 |

## Format

Parquet. Main columns:

| Column | Type | Description |
|---|---|---|
| `experiment` | string | Experiment name (RBP_cell line) |
| `sequence` | string | 101 nt RNA sequence |
| `icshape` | string | Per-base icSHAPE structure scores |
| `labels` | int64 | Binding label (0/1) |
| `inputs_embeds` | list<list<float64>>[101] | Numeric input combining sequence and structure |

The other 8 columns are original source fields kept under their original names.

## Usage

Example with [DNALLM](https://github.com/zhangtaolab/DNALLM): subset `experiments`, binary classification, `num_labels=2`.

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

ds = MsDataset.load("User040218/PrismNet", subset_name="experiments")
# tokenizer: the tokenizer of your DNALLM model; set max_length to match its tokenization
data = DNADataset(ds, tokenizer=tokenizer, max_length=103)
```

Select a single experiment before training, for example `ds.filter(lambda r: r["experiment"] == name)`.

## Notes

- Training on several RBPs together is not the task setting of the paper; use one experiment at a time.

## Source

If you use this dataset, please cite:

- Sun L, Xu K, Huang W, et al. Predicting dynamic cellular protein–RNA interactions by deep learning using in vivo RNA structures. Cell Research 2021;31(5):495-516. doi:10.1038/s41422-021-00476-y

## License

Declared license: MIT. The terms of the original data providers listed in Source also apply.