CYPMol: A Single Model Framework Integrating Functional Residues with Protein Features and Molecule Embeddings to...

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Cytochrome P450 (CYP) enzymes are crucial for metabolizing xenobiotics in humans.
• Accurate prediction of CYP-molecule interactions is vital for efficient drug development.
• Current models often neglect functional residues critical for CYP activity.

Purpose of the Study:

• To develop CYPMol, a deep learning framework for predicting CYP substrates, inhibitors, and bonds of metabolism (BoMs).
• To integrate functional residue information, protein language features, and small molecule embeddings into a unified model.
• To enhance the accuracy of predicting CYP-molecule interactions and metabolic pathways.

Main Methods:

• Developed a deep learning framework, CYPMol.
• Integrated functional residue information, protein language features, and pretrained small molecule embeddings.
• Trained and evaluated the model on CYP substrate, inhibitor, and BoM prediction tasks.

Main Results:

• CYPMol significantly outperformed state-of-the-art models in predicting human CYP substrates (MCC = 0.819) and inhibitors (MCC = 0.725).
• The framework successfully predicted specific BoMs for 537 diverse CYPs from various species.
• Incorporation of functional residue data improved predictive performance.

Conclusions:

• CYPMol offers a powerful and accurate framework for modeling CYP activity and interactions with small molecules.
• The model supports advancements in protein engineering and accelerates drug development processes.
• Publicly accessible data and framework facilitate further research and application.