A deep learning framework for residue-level prediction of protein properties, including:
- Intrinsically Disordered Regions (IDRs)
- Protein-binding residues
- RNA-binding residues
- DNA-binding residues
FlexBind is a sequence-based model designed to handle:
- Multi-scale interaction patterns along protein sequences
- Structural organization of functional residues
- Severe class imbalance in residue-level prediction tasks
The framework leverages pretrained protein embeddings and multi-scale representations to model diverse interaction behaviors.
FlexBind/
│
├── model.py
├── environment.yaml
├── README.md
│
├── DP81/
├── DP93/
├── DP94/
conda env create -f environment.yaml
conda activate flexbind
Run model (example):
python model.py
This repository uses DisProt-derived datasets:
- DP81
- DP93
- DP94