Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- protein
|
| 5 |
+
- structural-biology
|
| 6 |
+
- thermodynamic-stability
|
| 7 |
+
---
|
| 8 |
+
|
| 9 |
+
# DeepEF Datasets
|
| 10 |
+
|
| 11 |
+
Protein structure datasets used for training and evaluating [DeepEF](https://github.com/shaharec/DeepEF) —
|
| 12 |
+
a deep learning framework for predicting protein thermodynamic stability.
|
| 13 |
+
|
| 14 |
+
## Contents
|
| 15 |
+
|
| 16 |
+
| Path | Size | Description |
|
| 17 |
+
|------|------|-------------|
|
| 18 |
+
| `casp12_data_30/` | ~110GB | CASP12 structures with ProtT5 embeddings; per-protein `.pt` tensors split into train/test/valid-* |
|
| 19 |
+
| `Processed_K50_dG_datasets/` | ~2.2GB | K50 ddG mutation datasets with AlphaFold PDB models |
|
| 20 |
+
| `megascale_proteins.csv` | small | Megascale protein list |
|
| 21 |
+
|
| 22 |
+
## Data Format (casp12_data_30)
|
| 23 |
+
|
| 24 |
+
Each protein is a folder under `train/`, `test/`, or `valid-*/`:
|
| 25 |
+
```
|
| 26 |
+
PROTEIN_ID/
|
| 27 |
+
├── crd_backbone.pt # Backbone coordinates [seq_len, 4, 3]
|
| 28 |
+
├── ang.pt # Dihedral angles
|
| 29 |
+
├── mask.pt # Valid residue mask
|
| 30 |
+
├── seq_one_hot.pt # One-hot amino acid encoding [seq_len, 20]
|
| 31 |
+
├── seq.pt # Raw sequence string
|
| 32 |
+
├── seq_mut.pt # Mutant sequence string
|
| 33 |
+
├── proT5_emb.pt # ProtT5 embeddings [seq_len, 1024]
|
| 34 |
+
├── proT5_emb_cycle1-4.pt # Cyclic permutation embeddings
|
| 35 |
+
└── proT5_emb_mut.pt # Mutant ProtT5 embeddings
|
| 36 |
+
```
|
| 37 |
+
|
| 38 |
+
## Usage
|
| 39 |
+
|
| 40 |
+
```python
|
| 41 |
+
from huggingface_hub import snapshot_download, hf_hub_download
|
| 42 |
+
import torch
|
| 43 |
+
|
| 44 |
+
# Download a single protein
|
| 45 |
+
path = hf_hub_download(
|
| 46 |
+
repo_id="shaharec/deepef-data",
|
| 47 |
+
repo_type="dataset",
|
| 48 |
+
filename="casp12_data_30/train/1A0C_1_A/crd_backbone.pt",
|
| 49 |
+
)
|
| 50 |
+
coords = torch.load(path)
|
| 51 |
+
|
| 52 |
+
# Download an entire split (large!)
|
| 53 |
+
snapshot_download(
|
| 54 |
+
repo_id="shaharec/deepef-data",
|
| 55 |
+
repo_type="dataset",
|
| 56 |
+
allow_patterns="casp12_data_30/train/*",
|
| 57 |
+
local_dir="./data",
|
| 58 |
+
)
|
| 59 |
+
```
|