(M-025) Enhancing Model Interoperability and Workflow Integration in Clinical Trial Simulations​

Matthieu Coudron – Scientific Software Engirneering – Nova In Silico; Jean Ponchon – Scientific Software Engineering – Nova In Silico; Jérémy Villard – Scientific Software Engineering – Nova In Silico

Objectives: Reproducibility is critical in pharmacometric research, yet the diversity of modeling tools and formats creates significant barriers. Although PharmML has emerged as a promising standard for model exchange, its adoption remains limited across the industry. This study outlines how we leverage modern software engineering practices and large language models (LLMs) to accelerate the importation and exportation of both open standards such as SBML and commonly used industry formats. This approach mitigates the challenges posed by incomplete standard adoption without introducing new proprietary formats.

Methods: To support model interoperability, we implemented dedicated parsers and exporters tailored to widely used tools such as Monolix, NONMEM, and SimBiology. The development of these components was accelerated through the use of LLMs, which contributed to rapid prototyping and validation of syntax-specific rules and edge case handling for each target format. This approach allowed the team to scale parser implementation efficiently while ensuring correctness and completeness.

Model import functionality includes support for SBML, SimBiology’s Excel exports, and native Monolix and NONMEM projects, including population variability captured through virtual population designs. Export capabilities allow users to generate models in SBML, Julia, or R formats, promoting downstream compatibility and reproducibility.

A key enabler is a comprehensive and extensible API, which allows programmatic interaction with every component of the model, including structural equations, parameter definitions, and trial designs. This API supports advanced use cases such as automation, external pipeline integration, and fine-grained model editing. It is supplemented by a Python SDK and curated examples to help users integrate Jinkō into existing workflows.

To ensure robustness, we include a continuously running suite of thousands of automated tests covering all import/export operations. These include the SBML test suite (1819 models), of which 1391 are currently supported, and 1067 curated models from the BioModels database, with an 80% successful import rate. These systematic validations help prevent regressions and quantify format coverage over time.

Results: The implementation of these features has enhanced the platform's interoperability, allowing for model exchange and integration across various tools and workflows. Users have reported improved efficiency in model development and analysis, citing the ease of importing models and exporting them for use in different computational environments, reducing days of repetitive work to just minutes of validation. Hybrid development with LLMs cut implementation time by 50%, while automated testing and CI ensure code validity.

Conclusions: By combining robust import/export capabilities, comprehensive programmatic access, LLM-accelerated development, and disciplined software engineering practices, the Jinkō platform facilitates seamless integration of pharmacometric models into diverse analytical workflows. These enhancements support more efficient and flexible clinical trial simulations, ultimately accelerating drug development processes.

Citations: [1] Swat, M J et al. “Pharmacometrics Markup Language (PharmML): Opening New Perspectives for Model Exchange in Drug Development.” CPT: pharmacometrics & systems pharmacology vol. 4,6 (2015): 316-9. doi:10.1002/psp4.57
[2] Bizzotto, R et al. “PharmML in Action: an Interoperable Language for Modeling and Simulation.” CPT: pharmacometrics & systems pharmacology vol. 6,10 (2017): 651-665. doi:10.1002/psp4.12213
[3] Terranova, Nadia et al. “The Standard Output: A Tool-Agnostic Modeling Storage Format.” CPT: pharmacometrics & systems pharmacology vol. 7,9 (2018): 543-546. doi:10.1002/psp4.12339
[4] Kirouac, Daniel C et al. “Reproducibility of Quantitative Systems Pharmacology Models: Current Challenges and Future Opportunities.” CPT: pharmacometrics & systems pharmacology vol. 8,4 (2019): 205-210. doi:10.1002/psp4.12390
[5] BioModels repository: https://www.ebi.ac.uk/biomodels/
[6] Jinkō API Documentation: https://doc.jinko.ai/
[7] Jinkō API Helpers Python SDK: https://github.com/novainsilico/jinko-api-helpers-python
[8] Jinkō API Cookbook: https://github.com/novainsilico/jinko-api-cookbook

Keywords: Reproducibility, Model Interoperability, Programmatic API Access