{
  "id": 13037341,
  "title": "Hybrid Mechanistic-Neural Modeling of Concentration-Time Dynamics Generalizes Human Pharmacokinetics Prediction Across Unseen Chemical Space",
  "url": "https://urgent.news/2026/10/08/hybrid-mechanistic-neural-modeling-of-concentration-time-dynamics",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-08T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.10.01.756018v1?rss=1"
  },
  "original_language": "en",
  "account": "Pharmacokinetics concentration-time profiles reveal drug metabolism and excretion (ADME) dynamics. Optimizing these profiles during drug discovery helps minimize late-stage failure, yet predicting human pharmacokinetics from chemical structure remains challenging. Machine learning models disregard continuous ADME dynamics, whereas physiologically based pharmacokinetics (PBPK) models necessitate inaccessible parameters and enforce rigid, improperly specified compartmental structures.\n\nIntroducing PK-MUSE (PharmacoKinetic hybrid Mechanistic Universal System Equation), a novel approach that enhances a two-compartment model with bounded, state- and time-dependent neural corrections addressing elimination and intercompartmental distribution. Molecular structure determines compound-specific behavior, while learned corrections fine-tune dynamic equations. Across various chemical scaffolds and time frames, PK-MUSE demonstrated superior performance with the lowest mean exposure-endpoint MdFE. The model outperformed state-of-the-art machine learning and deep learning compartmental models by reducing MdFE by 7.2%, 45.2%, and 26.2%, respectively. By investigating these corrections, researchers identified that distribution differences, rather than elimination dynamics, led to misspecification. PK-MUSE advances digital twin capabilities in drug discovery by combining mechanistic simulators with data-driven, constrained neural dynamics, all while preserving interpretability and generalizing to new chemotypes.",
  "summary": "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…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}