Hybrid Mechanistic-Neural Modeling of Concentration-Time Dynamics Generalizes Human Pharmacokinetics Prediction Across Unseen Chemical Space
- 1Northeastern University
Abstract
The pharmacokinetics concentration-time profile encodes ADME dynamics. Its optimization in drug discovery curbs late-stage attrition, yet forecasting human pharmacokinetics from chemical structure remains difficult. Machine learning models ignore continuous ADME dynamics, whereas physiologically based pharmacokinetics (PBPK) models demand unavailable parameters and impose rigid, misspecified compartmental structures. We present PK-MUSE (PharmacoKinetic hybrid Mechanistic Universal System Equation), augmenting a two-compartment model with bounded, state- and time-dependent neural corrections to elimination and intercompartmental distribution. Molecular structure parameterizes compound-specific behavior, while learned corrections refine dynamic equations. Across scaffold and temporal out-of-distribution evaluations, PK-MUSE achieved the lowest mean exposure-endpoint MdFE, reducing MdFE by 7.2%, 45.2%, and 26.2% versus state-of-the-art machine learning (HML- RF), deep learning compartmental (DeepCt), and PBPK (PK-Sim) models. Interrogating the corrections localized misspecification to distribution rather than elimination. By extending mechanistic simulators with data-driven constrained neural dynamics that preserve interpretability and generalize to unseen chemotypes, PK-MUSE advances digital twins for drug discovery.
Related
- Mechanistic PBPK Modeling of Antibody–Drug Conjugates: Integration of Antibody Disposition, Linker Deconjugation, and Payload Pharmacokinetics
- Formononetin as a Natural Product-Derived Candidate for Diabetic Microvascular Complications: Mechanistic Evidence, Pharmacokinetics, Formulation Strategies, and Translational Barriers
- Population pharmacokinetics modeling of albendazole and its active metabolite, albendazole sulfoxide in mass drug administration against lymphatic filariasis
- Impact of GPR39 Inhibition on Tissue Oxygen Tension in Acute Myocardial Infarction: Modeling Using an Indirect Response Pharmacokinetics/Pharmacodynamics Approach
- Physiologically Based Pharmacokinetic Modeling of Venetoclax in Pediatric Acute Myeloid Leukemia: Food Effects, Age-Related Pharmacokinetics, and Voriconazole-Mediated Drug-Drug Interactions.
BibTeX
@misc{badkul2026hybrid,
title = {{Hybrid Mechanistic-Neural Modeling of Concentration-Time Dynamics Generalizes Human Pharmacokinetics Prediction Across Unseen Chemical Space}},
author = {A. Badkul and L. Cai and M. Kim and J. Wang and L. Xie},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.01.756018},
url = {https://www.biorxiv.org/content/10.64898/2026.10.01.756018v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Hybrid Mechanistic-Neural Modeling of Concentration-Time Dynamics Generalizes Human Pharmacokinetics Prediction Across Unseen Chemical Space
AU - A. Badkul
AU - L. Cai
AU - M. Kim
AU - J. Wang
AU - L. Xie
PY - 2026
DA - 2026/10/01
PB - bioRxiv
DO - 10.64898/2026.10.01.756018
UR - https://www.biorxiv.org/content/10.64898/2026.10.01.756018v1
AB - The pharmacokinetics concentration-time profile encodes ADME dynamics. Its optimization in drug discovery curbs late-stage attrition, yet forecasting human pharmacokinetics from chemical structure remains difficult. Machine learning models ignore continuous ADME dynamics, whereas physiologically based pharmacokinetics (PBPK) models demand unavailable parameters and impose rigid, misspecified compartmental structures. We present PK-MUSE (PharmacoKinetic hybrid Mechanistic Universal System Equation), augmenting a two-compartment model with bounded, state- and time-dependent neural corrections to elimination and intercompartmental distribution. Molecular structure parameterizes compound-specific behavior, while learned corrections refine dynamic equations. Across scaffold and temporal out-of-distribution evaluations, PK-MUSE achieved the lowest mean exposure-endpoint MdFE, reducing MdFE by 7.2%, 45.2%, and 26.2% versus state-of-the-art machine learning (HML- RF), deep learning compartmental (DeepCt), and PBPK (PK-Sim) models. Interrogating the corrections localized misspecification to distribution rather than elimination. By extending mechanistic simulators with data-driven constrained neural dynamics that preserve interpretability and generalize to unseen chemotypes, PK-MUSE advances digital twins for drug discovery.
ER -