Predictive Models for Human Cytochrome P450 3A7 Selective Inhibitors and Substrates

Area of Science:

• Pharmacology and Toxicology
• Computational Chemistry
• Drug Metabolism

Background:

• Prescription drug use poses unique risks to fetuses and neonates compared to adults.
• Cytochrome P450 (CYP) 3A subfamily expression changes significantly after birth (CYP3A7 to CYP3A4), impacting drug metabolism differently across age groups.

Purpose of the Study:

• To develop novel predictive models for identifying inhibitors and substrates of CYP3A7 and CYP3A4 using chemical structure analysis.
• To enable distinct evaluation of drug safety and efficacy for fetal/neonatal and adult populations.

Main Methods:

• Utilized chemical structure analysis to build predictive models for CYP3A7/CYP3A4 inhibitors and substrates.
• Evaluated model performance using area under the receiver operating characteristic curve (AUC-ROC), balanced accuracy (BA), and Matthews correlation coefficient (MCC).
• Validated model predictions using known CYP3A7/CYP3A4 inhibitor/substrate compounds and identified significant structural features.

Main Results:

• Optimal models achieved AUC-ROC values ranging from 0.72 to 0.84, depending on the dataset and model type.
• Model performance varied based on dataset type, model, rebalancing method, and feature set.
• Identified key structural features associated with selective or common CYP3A7/CYP3A4 inhibition and substrate activity.

Conclusions:

• Developed validated predictive models for CYP3A7/CYP3A4 inhibitors and substrates, applicable to both fetal/neonatal and adult populations.
• These models serve as a tool for rapid safety and efficacy assessment of new drugs in distinct developmental stages.
• Identified structural features can guide the design and optimization of drugs for improved fetal/neonatal safety and efficacy.