---
license: mit
task_categories:
  - feature-extraction
language:
  - en
tags:
  - biology
  - proteins
  - thermophiles
  - protein-engineering
  - alphafold
size_categories:
  - n<1K
---

# APED: African Protein Engineering Dataset

A curated dataset of protein structures from African thermophilic organisms, designed for machine learning applications in protein engineering.

## Dataset Description

This dataset contains structural and sequence features extracted from AlphaFold-predicted structures of proteins from thermophilic organisms found in African extreme environments (hot springs, volcanic regions).

### Features (19 total)

| Feature | Description |
|---------|-------------|
| `uniprot_id` | UniProt accession |
| `organism` | Source organism |
| `sequence` | Amino acid sequence |
| `sequence_length` | Number of residues |
| `mean_plddt` | Mean AlphaFold confidence |
| `helix_fraction` | α-helix content |
| `sheet_fraction` | β-sheet content |
| `coil_fraction` | Coil content |
| `hydrophobicity` | Mean hydrophobicity |
| `charge` | Net charge at pH 7 |
| `molecular_weight` | Molecular weight (Da) |
| `isoelectric_point` | Predicted pI |
| `instability_index` | Sequence instability |
| `aromaticity` | Aromatic residue fraction |
| And more... | |

### Statistics

- **500 proteins** with full features
- **981 AlphaFold structures** downloaded
- **3 novel designs** with pLDDT > 90%

## Novel Protein Designs

Included are 3 computationally designed proteins generated using:
1. **RFdiffusion** - backbone generation
2. **ProteinMPNN** - sequence design  
3. **AlphaFold** - structure validation

Best design: **91.9% pLDDT, 0.74Å RMSD**

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("electricsheepafrica/APED-African-Protein-Engineering-Dataset")
```

## Citation

If you use this dataset, please cite:
```
@dataset{aped2024,
  title={APED: African Protein Engineering Dataset},
  author={Electric Sheep Africa},
  year={2024},
  publisher={HuggingFace}
}
```

## License

MIT License