| --- |
| title: TREAD Protein Repeat Annotation |
| emoji: 🧬 |
| colorFrom: indigo |
| colorTo: blue |
| sdk: gradio |
| sdk_version: 5.49.1 |
| python_version: 3.10.13 |
| app_file: app.py |
| pinned: false |
| --- |
| |
| # TREAD Protein Repeat Annotation |
|
|
| Interactive Gradio web demo for the pretrained TREAD models from: |
|
|
| - https://github.com/KYQiu21/TREAD |
| - ProtT5 encoder: `Rostlab/prot_t5_xl_half_uniref50-enc` |
|
|
| The Space exposes the two inference workflows used in the public TREAD Colab demo: |
|
|
| 1. RepeatsDB repeat annotation (segmentation + six repeat-fold heads) |
| 2. Beta-propeller blade annotation |
|
|
| ## Compute |
|
|
| This version is designed for a **ZeroGPU Gradio Space** and offers two execution paths: |
|
|
| - **Run with ZeroGPU (recommended):** dynamically requests a GPU only during inference. |
| - **Run on CPU (fallback):** does not require GPU quota, but ProtT5-XL inference is much slower. |
|
|
| ## Reproducibility |
|
|
| The app fetches the same fixed TREAD Git commit used by the working Colab demo: |
|
|
| `7a7b3f571778cb89035c798c88df7eacb63ed2c1` |
|
|
