A Time-Embedding Network Models the Ontogeny of 23 Hepatic Drug Metabolizing Enzymes

Area of Science:

• Pharmacology
• Computational Biology
• Pediatric Medicine

Background:

• Pediatric patients face higher risks of adverse drug reactions due to limited safety data.
• Age-dependent changes in drug absorption, distribution, metabolism, and excretion complicate pediatric risk assessment.
• The ontogeny of drug metabolism enzymes significantly impacts age-dependent drug toxicity.

Purpose of the Study:

• To develop a computational model for predicting age-related variations in drug metabolism enzyme expression.
• To assess the utility of time-embedding neural networks in modeling enzyme ontogeny.
• To estimate age-dependent reactive metabolite exposure and identify potential toxicity mechanisms in pediatric populations.

Main Methods:

• Implementation of time-embedding neural networks to model the ontogeny of 23 drug metabolism enzymes.
• Utilizing the network to capture population-level variations in enzyme expression as a function of age.
• Combining the ontogeny model with additional data to estimate age-dependent reactive metabolite exposure.

Main Results:

• The time-embedding network accurately modeled the ontogeny of 23 drug metabolism enzymes.
• The model successfully recapitulated known demographic factors influencing CYP3A5 expression.
• It effectively captured nonlinear dynamics of CYP2D6 expression, outperforming standard neural networks.
• Age-dependent changes in reactive metabolite exposure for valproic acid and dextromethorphan were identified.

Conclusions:

• Time-embedding neural networks provide a robust method for modeling age-dependent drug metabolism enzyme expression in pediatric populations.
• This approach enhances the estimation of reactive metabolite exposure, aiding in the evaluation of drug toxicity risks.
• The findings suggest potential mechanisms for valproic acid toxicity and offer a valuable tool for pediatric drug safety research.