PK/PD Modeling for CNS drug development

This case study shows how integrated PK/PD modeling with machine learning supported compound selection in a complex CNS program involving delayed efficacy and heterogeneous preclinical studies. Using Phoenix NLME, a unified mechanistic framework enabled objective cross compound comparisons and translational decision making. The analysis demonstrated that the most potent compound was not necessarily the optimal choice for first in human studies, leading to confident selection of the candidate with the most favorable overall pharmacological profile.

ソフトウェア

Phoenix® NLME

ソリューション

Preclinical, Early Clinical

The challenge

The project aimed to support final compound selection for first-in-human studies within a CNS drug development program characterized by complex and indirect pharmacology. The system involved a biomarker with rapid turnover, a slowly developing efficacy marker, and compounds acting through inhibition of biomarker synthesis, resulting in delayed efficacy responses. Data were pooled across multiple heterogeneous preclinical study designs—including acute oral dosing with rich PK and biomarker sampling, repeated-dose efficacy studies, and constant-rate infusion studies using osmotic mini-pumps – across 29 compounds with diverse pharmacological properties. The variability across compounds and study designs, combined with the need to consistently integrate pharmacokinetics, biomarker dynamics, efficacy development, and both in vitro (Kd) and in vivo (IC50) potency information, made compound selection highly complex and required a unified, unbiased analytical framework.

The solution

A machine learning–enabled nonlinear mixed-effects modeling strategy was implemented within Certara’s platform to simultaneously analyze all studies using mechanistic PK/PD models. The framework explicitly separated drug-specific parameters (e.g., potency) from physiological system parameters (e.g., baselines and turnover rate constants) while capturing binding kinetics, biomarker turnover, and delayed efficacy responses. By integrating in vitro and in vivo potency information within a single coherent modeling structure, the approach enabled consistent cross study and cross compound comparisons and strengthened PK/PD Modeling for CNS drug development.

The impact

The integrated analysis provided a robust, objective basis for compound selection and translational decision-making. Importantly, it demonstrated that the most in vivo potent compound is not necessarily the optimal candidate for human studies, thereby improving confidence in selecting the compound with the most favorable overall pharmacological profile for first-in-human development.

This compound, referred to as the “ugly duckling”, was ultimately selected for first time in human studies.

Overall, the case study illustrates how Phoenix NLME enables objective, system level decision making, supports confident compound selection, and accelerates translational insight in complex PK/PD scenarios.