Pavel Balazki 1, Stephan Schaller 1
1 ESQlabs GmbH (Saterland , Germany)
Objective
Current modelling workflows in pharmacometrics and systems pharmacology rely heavily on manual steps, particularly when assembling standard model structures and integrating heterogeneous datasets from multiple sources. Tasks such as importing compound-specific physicochemical properties, entering in vivo and in vitro study parameters, and manually configuring model components within simulation software are time-consuming, error-prone, repetitive, and thus not scalable. These manual bottlenecks consume substantial modelling resources, especially during high-throughput screening or when evaluating complex scenarios across large portfolios. Furthermore, the heterogeneity of data formats and units, combined with human transcription and configuration errors, increases the likelihood of inconsistencies and reduces reproducibility across projects and teams.
To address these challenges, we propose an extensible framework that automates the assembly of physiologically based pharmacokinetic (PBPK) and quantitative systems pharmacology (QSP) models for the Open Systems Pharmacology (OSP) platform (1). The framework integrates standardized databases for compound data and in vivo study information, along with modular QSP components, enabling a streamlined, scalable, and fully traceable modelling process that supports the entire MIDD cycle from discovery to regulatory submission.
Methods
The framework leverages the OSP Suite, an open-source ecosystem comprising PK-Sim® and MoBi®, and extends the capabilities of the {ospsuite} R package. The core innovation lies in the programmatic manipulation of OSP Project Snapshots: JSON-based model definitions that describe complete PBPK and QSP projects.
We developed a novel R interface, MoBi.R, which enables direct interaction with MoBi projects. This interface allows for the automatic generation of new simulations by merging existing PBPK templates with QSP extension modules. The workflow architecture consists of three layers:
1. Data Layer: Compound information (physicochemical properties), in vivo study results (PK profiles), and metadata are curated in a relational database following standardized schemas.
2. Module Layer: QSP extension modules (e.g., specific organ models, signaling pathways, or disease progression models) are organized and version-controlled in GitHub repositories.
3. Assembly Layer: The MoBi.R engine queries the database, retrieves the necessary QSP modules as PKML files, and programmatically injects specific parameters, initial conditions, and structural modules into a base PBPK model structure defined via Snapshots.
The framework supports additional interfaces in R or Python, enabling connections to proprietary or public third-party databases without altering the underlying workflow logic.
Results
We successfully implemented an automated, database-driven MIDD workflow that supports the efficient construction, updating, and qualification of complex PBPK and PBPK–QSP models without manual GUI interaction.
Automated Assembly and Scalability: By querying the centralized database for compound properties and study protocols, the MoBi.R interface demonstrated the ability to automatically generate full PBPK modelling projects. In high-throughput applications, this framework successfully processed batches of compounds, reducing the time required to assemble and configure verified PBPK models from hours and days of manual work to minutes of computational time, while ensuring 100% consistency between source data and model parameters.
Modular QSP Extension: To demonstrate the framework’s capability to handle complex systems pharmacology structures, we applied the workflow to a Thyroid Disruption Quantitative Systems Toxicology (QST) platform [2]. The automated workflow successfully extended a standard whole-body PBPK model by injecting a detailed thyroid organ module sourced directly from a GitHub repository. The MoBi.R interface automatically mapped the relevant molecular entities (e.g., T3, T4, TSH) and extended the initial conditions building blocks.
Conclusion
Overall, the proposed framework accelerates initial model setup and reduces manual burden, while promoting standardization in terminology, modelling methodology, and documentation practices. By enabling automated assembly of complex PBPK QSP structures, the workflow eliminates operational error sources in collaborative model development and supports rapid scaling of modelling activities across projects and therapeutic areas. The enforced structure ensures traceability, reproducibility, and transparent scientific communication, ultimately enhancing the reliability and regulatory acceptability of complex PBPK and QSP models.
References:
1. Community OSP. Open Systems Pharmacology [Internet]. [cited 2024 Mar 11]. Available from: https://www.open-systems-pharmacology.org/
2. Physiology-based systems pharmacology model of thyroid hormones regulation in rat and human. (accessed 02.2026) https://github.com/Open-Systems-Pharmacology/Thyroid-Hormones-PB-QSP-Model
Reference: PAGE 34 (2026) Abstr 12063 [www.page-meeting.org/?abstract=12063]
Poster: Drug/Disease Modelling - Other Topics