Machine learning-based prediction of pharmacokinetic parameters for individualized drug dosage optimization

This study presents a data-driven approach to predict pharmacokinetic parameters and generate concentration–time curves for a two-compartment model. The method employs inverse modelling using optimization algorithms to estimate patient-specific parameters from observed data. Machine learning techniques are then applied to solve the forward problem, enabling the prediction of concentration–time profiles for various dose levels. The study incorporates patient background characteristics to improve predictive performance, aiming to enable individualized drug dosing. Results demonstrate accurate parameter prediction and close matching of generated curves to observed data across six dose levels. This approach offers a novel framework for personalizing pharmacokinetic profiles and improving drug dosing strategies and therapeutic outcomes in clinical practice.