DEEPCYPs

DEEPCYPs predicts inhibitory activities of small molecules against cytochrome P450 isoforms 1A2, 2C9, 2C19, 2D6, and 3A4 to support assessment of drug metabolism and drug–drug interaction risk.

Key Features:

• FP-GNN model: Uses the FP-GNN (Fingerprint Graph Neural Network) deep learning architecture to model molecular features relevant to CYP inhibition.
• Target isoforms: Predicts inhibitory activity specifically for CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4.
• In silico classification: Implements in silico classification models that analyze molecular structures and their potential interactions with CYP enzymes.
• Performance metrics: Reports average Area Under the Curve (AUC) 0.905, F1 score 0.779, Balanced Accuracy (BA) 0.819, and Matthews Correlation Coefficient (MCC) 0.647.
• Robustness validation: Employs Y-scrambling tests to validate that predictive performance is not due to chance correlations.
• Interpretability: Identifies critical structural fragments associated with CYP inhibition to provide mechanistic insights.

Scientific Applications:

• Drug–drug interaction assessment: Predicts CYP inhibition profiles to help assess potential DDIs during drug development.
• Early-stage drug discovery: Identifies and prioritizes potential CYP inhibitors for medicinal chemistry optimization.
• Pharmacokinetics and safety: Supports evaluation of drug metabolism and safety profiling related to CYP-mediated biotransformation.
• Xenobiotic and endogenous metabolism analysis: Assesses interactions of xenobiotics and endogenous compounds with key CYP isoforms.

Methodology:

FP-GNN (Fingerprint Graph Neural Network)-based in silico classification models analyze molecular structures for CYP inhibition, with performance evaluated by AUC, F1, Balanced Accuracy, MCC and validated using Y-scrambling.