Mariia Myshkina 1, Jan Grzegorzewski 1, Matthias König 1,2
1 Humboldt-Universität zu Berlin (Berlin, Germany), 2 University of Stuttgart (Stuttgart, Germany)
Introduction: The rapid growth of published pharmacokinetics (PK) and pharmacodynamics (PD) data from clinical and preclinical studies requires structured, interoperable, and reproducible data infrastructures. In line with FAIR principles [1], pharmacometric data must be standardized, contextualized, and comparable across studies. We previously developed PK-DB [2], an open database for curated PK/PD data with detailed study and cohort metadata. While successfully used in modeling workflows, the original system was not designed for AI-driven analyses or digital twin applications.
Objectives: PK-DB 2.0 aims to provide an open, standardized, and computationally accessible PK/PD platform tailored to pharmacometric modeling, automated workflows, artificial intelligence, and digital twin development, with a focus on data quality, interoperability, and machine-readability.
Methods: PK-DB 2.0 implements a unified data model covering study design, interventions, concentration–time profiles, derived PK parameters, and pharmacodynamic endpoints. Data are curated through collaborative workflows with strong validation rules, automated quality checks, unit normalization, explicit representation of experimental uncertainties, and ontology-based semantic annotation.
A redesigned REST API provides fast, reliable, bi-directional programmatic access independent of programming language. Importantly, PK-DB 2.0 introduces a dedicated Model Context Protocol (MCP) server, enabling structured, automated access by AI clients and seamless integration into AI-driven modeling and digital twin workflows.
Results: PK-DB 2.0 currently contains curated data from more than 700 clinical studies, substantially expanding previous releases. Key advances include extended support for pharmacodynamic endpoints (e.g., coagulation, blood pressure, anthropometric measures), ontology-based annotation of biological and clinical entities, strong structural validation rules, and an AI-ready access layer via REST and MCP. These features enable direct, reproducible integration of curated PK/PD datasets into population modeling, PBPK modeling, machine learning pipelines, and digital twin frameworks.
Conclusions: PK-DB 2.0 provides a unified, machine-readable PK/PD data infrastructure for modern pharmacometrics. By combining rigorous curation, standardized representation, and explicit AI integration through a dedicated MCP server, the platform supports reproducible model development, cross-study analyses, and next-generation AI- and Digital Twin–enabled decision support.
References:
[1] Wilkinson, M.D. et al. 2016. The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data. 3, (Mar. 2016), 160018. DOI:https://doi.org/10.1038/sdata.2016.18.
[2] Grzegorzewski, J. et al. 2021. PK-DB: pharmacokinetics database for individualized and stratified computational modeling. Nucleic Acids Research. 49, D1 (Jan. 2021), D1358–D1364. DOI:https://doi.org/10.1093/nar/gkaa990.
Reference: PAGE 34 (2026) Abstr 12097 [www.page-meeting.org/?abstract=12097]
Poster: Drug/Disease Modelling - Absorption & PBPK