ProtScout is a Python package that enables ranking of protein sequences based on multiple properties predicted by state-of-the-art AI models. It provides a unified interface to assess and compare proteins using various characteristics such as stability, solubility, and catalytic efficiency.
- 🧬 Comprehensive protein property prediction
- 🤖 Integration with multiple AI models
- 📊 Flexible ranking system with customizable weights
- 🚀 Efficient batch processing
- 📈 Property visualization tools
- 🔄 Combined prediction support
pip install protscoutOr install from source using Poetry:
git clone https://github.com/Robaina/ProtScout
cd protscout
poetry installfrom protscout import ProteinRanker
# Initialize a combined predictor with custom weights
ranker = ProteinRanker(
predictors=[
KineticPredictor(substrate_smiles="CC(=O)Oc1ccc([N+](=O)[O-])cc1", weight=1.0),
ThermostabilityPredictor(weight=0.5),
SolubilityGATSolPredictor(weight=0.5)
]
)
# Load sequences
sequences = ["MAEGEITTFTALTEKFNLPPGNYKKPKLLYCSNG", ...]
# Get combined rankings
rankings = ranker.rank(sequences)- KineticPredictor: Predicts enzyme kinetic parameters (kcat, KM, Ki)
- KMpredictor: Specialized predictor for Michaelis constants (KM)
- AffinityGNINApredictor: Predicts protein-ligand binding affinity
- DeepStabPtmPredictor: Predicts protein melting temperature (Tm)
- ThermostabilityPredictor: Predicts protein thermostability using TemBERTure
- GeoPocPredictor: Predicts optimal temperature, pH, and salt conditions
- SolubilityGATSolPredictor: Predicts protein solubility using GATSol
- EpHodPredictor: Predicts optimal pH conditions
- CombinedPredictor: Enables weighted combination of multiple predictors
| Predictor | Input Requirements | Properties Predicted | Model Base |
|---|---|---|---|
| KineticPredictor (CatPred) | Sequences, SMILES | kcat, KM, Ki | esm2_t33_650M_UR50D |
| KMpredictor | Sequences, SMILES | KM | ESM-2 |
| AffinityGNINApredictor | Sequences, PDB files, SDF | Binding Affinity | GNINA |
| DeepStabPtmPredictor | Sequences | Melting Temperature | Prot-T5-XL |
| ThermostabilityPredictor | Sequences | Thermostability | ProtBERT |
| GeoPocPredictor | Sequences, (optional PDBs) | Temperature, pH, Salt | esm2_t36_3B_UR50D |
| SolubilityGATSolPredictor | Sequences, PDB files | Solubility | ESM-1b |
| EpHodPredictor | Sequences | Optimal pH | ESM-1v |
- Optimize containers to enable input of protein embeddings for ESM-2 models
- Standardize embedding input across all models
- Support direct embedding input for ESM-based models
- Unify embedding computation to avoid redundant calculations
- Add support for custom model weights
- Implement caching for embeddings
- Add batch prediction support for all predictors
For detailed documentation, visit protscout.readthedocs.io
We welcome contributions! Please see our Contributing Guidelines for details.
This project is licensed under the GPL-3.0 License - see the LICENSE file for details.
- Thanks to all the teams behind the integrated AI models
- Contributors and maintainers of the essential dependencies
- The computational biology community for valuable feedback