Robust Adaptation of Pre-trained Models to Inaccurate Biological Labels via Weak Supervision

🚀 Introduction (VenusWSL)

We present VenusWSL, a weakly supervised learning framework that addresses label noise in protein property prediction by leveraging a Gaussian Mixture Model to separate clean and noisy labels, enabling more robust training through a teacher-student approach.

📑 Results

News

Paper Results

🛫 Requirement

Conda Environment

Please make sure you have installed Anaconda3 or Miniconda3.

conda env create -f environment.yaml
conda activate protein

Hardware

We recommend using a GPU with at least 12GB memory.

🧬 Start with VenusWSL

Get PLM Embedding

bash script/get_plm_embed.sh

Prepare Dataset

bash script/prepare_dataset.sh

Train with VenusWSL

bash script/train.sh

🙌 Citation

if you find this work useful, please cite:

📝 License

This project is licensed under the terms of the CC-BY-NC-ND-4.0 license.