metadata

title: Transeption IGEM BASISCHINA 2025
emoji: 🧬
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.34.2
app_file: app.py
pinned: false
license: mit
suggested_hardware: zero-a10g
models:
  - PascalNotin/Tranception_Small
  - PascalNotin/Tranception_Medium
  - PascalNotin/Tranception_Large

Tranception Protein Fitness Prediction - BASIS-China iGEM 2025

Welcome to BASIS-China iGEM Team's deployment of Tranception on Hugging Face Spaces!

About This Project

This is an implementation of the Tranception model for protein fitness prediction, deployed by the BASIS-China iGEM Team 2025. Our goal is to make advanced protein engineering tools accessible to the synthetic biology community.

Features

In silico directed evolution: Iteratively improve protein fitness through single amino acid substitutions
Comprehensive fitness analysis: Generate heatmaps showing fitness scores for all possible mutations
Zero GPU support: Leverages Hugging Face's dynamic GPU allocation for efficient inference
Multiple model sizes: Choose between Small, Medium, and Large models based on your needs

Technical Implementation

This deployment utilizes Hugging Face's Zero GPU infrastructure, which:

Dynamically allocates H200 GPU resources when available
Seamlessly falls back to CPU processing when GPUs are unavailable
Ensures efficient resource management for all users

About BASIS-China iGEM Team

We are a high school synthetic biology team participating in the International Genetically Engineered Machine (iGEM) competition. Our 2025 project focuses on protein engineering and computational biology applications.

Credits

This implementation is based on: Tranception: Protein Fitness Prediction with Autoregressive Transformers and Inference-time Retrieval by Pascal Notin, Mafalda Dias, Jonathan Frazer, Javier Marchena-Hurtado, Aidan N. Gomez, Debora S. Marks, and Yarin Gal.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference