---
pretty_name: DisProt Protein Disorder Annotations
license: cc-by-4.0
tags:
- biology
- proteins
- intrinsic-disorder
- disprot
- annotation
- parquet
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*.parquet
  - split: test
    path: data/test-*.parquet
---

# DisProt Protein Disorder Annotations

DisProt is a manually curated database of intrinsically disordered proteins and regions, including experimental evidence and functional annotations for protein disorder.

## Splits

The split is deterministic by DisProt ID: `sha256(disprot_id) % 10`. Bucket `0` is `test`; buckets `1` through `9` are `train`.

| Split | Rows |
|---|---:|
| train | 2,875 |
| test | 324 |
| total | 3,199 |

## Dataset Statistics

| Field | Value |
|---|---:|
| Protein entries | 3,199 |
| Curated region rows | 13,396 |

## Usage

```bash
pip install datasets
```

Load all splits:

```python
from datasets import load_dataset

ds = load_dataset("LiteFold/DisProt")
print(ds)
print(ds["train"][0]["disprot_id"])
```

Load one split:

```python
from datasets import load_dataset

train = load_dataset("LiteFold/DisProt", split="train")
```

Filter proteins with high disorder content:

```python
from datasets import load_dataset

ds = load_dataset("LiteFold/DisProt", split="train")
high_disorder = ds.filter(lambda row: row["disorder_content"] is not None and row["disorder_content"] >= 0.5)
print(high_disorder[0])
```

Load the region-level metadata:

```python
import pandas as pd
from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="LiteFold/DisProt",
    repo_type="dataset",
    filename="metadata/regions.parquet",
)
regions = pd.read_parquet(path)
print(regions.head())
```

## Key Columns

| Column | Description |
|---|---|
| `disprot_id` | DisProt protein identifier. |
| `accession` | UniProt accession. |
| `name` | Protein name. |
| `organism` | Organism name. |
| `ncbi_taxon_id` | NCBI taxonomy ID. |
| `length` | Protein sequence length. |
| `disorder_content` | Fraction of sequence annotated as disordered. |
| `dataset_labels` | DisProt dataset labels. |
| `sequence` | Protein sequence. |
| `region_count` | Number of curated region annotations. |
| `region_ids` | Curated region IDs. |
| `region_starts` | Region start positions. |
| `region_ends` | Region end positions. |
| `region_terms` | Region term names. |
| `evidence_codes` | Region evidence codes. |
| `reference_ids` | Region reference IDs. |
| `cross_refs` | Region cross-references such as PDB IDs. |
| `feature_databases` | Feature sources such as Pfam or Gene3D. |
| `feature_ids` | Feature IDs. |
| `gene_names` | Gene names and synonyms. |
| `split_bucket` | Deterministic split bucket from `sha256(disprot_id) % 10`. |

# Citation

```
@article{quaglia2022disprot,
  title   = {{DisProt} in 2022: improved quality and accessibility of protein intrinsic disorder annotation},
  author  = {Quaglia, Federica and M{\'e}sz{\'a}ros, B{\'a}lint and Salladini, Elisa and others},
  journal = {Nucleic Acids Research},
  volume  = {50},
  number  = {D1},
  pages   = {D480--D487},
  year    = {2022},
  doi     = {10.1093/nar/gkab1082}
}
```