README.md

metadata

title: AllerTrans
emoji: 🔥
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 5.23.3
app_file: app.py
pinned: false
license: mit
short_description: A deep learning method for protein allergenicity prediction
models:
  - Rostlab/prot_t5_xl_uniref50
  - facebook/esm2_t33_650M_UR50D
tags:
  - biology
  - protein-language-models
  - esm
  - prott5
  - allergen-prediction
  - bioinformatics
  - deep-learning
  - research
  - protein
  - protein-sequence

🧬 AllerTrans: a deep learning method for predicting the allergenicity of protein sequences

Welcome to the interactive demo for AllerTrans, a state-of-the-art computational tool designed to predict protein allergenicity using advanced Protein Language Models (pLMs).

🔗 Read the Paper in Biology Methods and Protocols (Oxford Academic) 🔗 Access the Source Code on GitHub

---

🔬 About This Space

AllerTrans leverages the power of deep learning to accurately assess whether a given protein sequence is an allergen. This Space provides a user-friendly interface to test individual sequences quickly, allowing researchers to evaluate the model without needing to install complex dependencies or configure heavy machine learning environments locally.

By ensembling methods, AllerTrans achieves high sensitivity (97.69%), and specificity (97.80%) for sequence-based allergen prediction using Algpred public benchmark.

---

🤖 Foundational Models

This research builds upon the incredible foundational Protein Language Models developed by the open-science community. AllerTrans relies on:

• ProtT5-XL-UniRef50 developed by Rostlab
• ESM-2 developed by facebook research (Meta AI)

---

📝 Citation

If you use AllerTrans in your research or find our methodology helpful, please cite our published work:

@article{allertrans,
  title={AllerTrans: a deep learning method for predicting the allergenicity of protein sequences},
  author={Sarlakifar, Faezeh and others}, 
  journal={Biology Methods and Protocols},
  year={2025},
  publisher={Oxford University Press},
  doi={https://doi.org/10.1093/biomethods/bpaf040}
}