Victor Sokolov 1, Anna Mikhailova 1,2, Yaroslav Ugolkov 1,2, Anatoly Pokladyuk 1,3, Alina Melnikova 1, Kirill Zhudenkov 1,4
1 M&S Decisions (Dubai, United Arab Emirates), 2 Marchuk Institute of Numerical Mathematics of the Russian Academy of Sciences (INM RAS) (Moscow, Russia), 3 Sirius University (Sochi, Russia), 4 Research Center of Model-Informed Drug Development, Sechenov First Moscow State Medical University (Moscow, Russia)
Introduction
Pharmacometrics analysis can be represented as workflow comprising multiple highly specialized steps that typically include data processing, data digitization, data visualization, solving of differential equations, parameter estimation procedures, and reporting [1, 2]. The requirements of reproducibility and transparency motivate the use of scripting languages to execute these steps, and multiple R packages have been developed either to perform specialized tasks or to provide seamless integration of various functions into a unified pipeline, e.g. xpose [3], ggPMX [4], rxode2 [5], mrgsolve [6], etc. In this work, we present a novel R package, SimuRg – a comprehensive wrapper that enables a software-agnostic NLME modeling workflow.
Methods
SimuRg (v0.2.0) is an open-source R package that integrates multiple ODE-solving algorithms, APIs for NLME estimation tools (e.g., Monolix, NONMEM), and ggplot2-based visualization. The package operates on a standardized JSON object encapsulating parameter estimates, random effects, covariate definitions, and dataset references. SimuRg exports over 20 functions organized into four domains: (1) model building and fitting; (2) parameter extraction and model comparison; (3) goodness-of-fit (GOF) diagnostics; and (4) model simulation. The package is integrated into the Simurg ecosystem – a web-based platform built on TypeScript with applications capable of generating Simurg function calls with appropriate arguments, assembling them into R Markdown (Rmd) documents and refining by Sirin AI agent.
Results
A key capability of SimuRg is the ability to execute all steps of the model development and simulation workflow independently of the estimation software used. This is achieved through the sg_fit function, which generates control files for the Simurg ecosystem’s estimation algorithms or other conventional software (e.g., Monolix), executes the analysis, and produces a standardized JSON object organized into layers resembling SDTAB, PATAB, COTAB, and related structures. Structural model input is likewise software-agnostic: in the current version, models can be specified in either MLXTRAN or rxode2 format. Model selection and evaluation are further supported by sg_modbuild and sg_multistart. When sg_fit is not used for parameter estimation, sg_converter can generate the standardized JSON object from existing computational projects.
The resulting JSON object enables a suite of functions – including sg_parsum, sg_modcomp, sg_gof_obpr, sg_gof_par_dist, sg_gof_res_dist, sg_gof_tp, and others – to deliver a comprehensive set of model diagnostics.
Differential equations are solved via either rxode2 or the Simurg ecosystem’s engine, supporting simulation-based diagnostics such as VPCs and prediction distribution assessment, local and global sensitivity analyses, and forward simulation through the sg_sim function, compatible with both MLXTRAN and rxode2 formats. Additionally, SimuRg supports virtual population generation through UMAP-based data synthesis.
Conclusions
SimuRg provides a unified, software-agnostic framework covering the full NLME modeling workflow. Its standardized JSON-based architecture allows analyses to be performed independently of estimation engines, promoting efficiency, flexibility, and reproducibility, further enhanced by integration with the Simurg ecosystem and the Sirin AI agent. The package is under active development, with planned enhancements including NONMEM compatibility, Bayesian ODE-based modeling, and stand-alone model syntax translation.
References:
[1] Byon et al. CPT Pharmacometrics Syst Pharmacol. 2013.
[2] Sokolov et al. Handbook of Experimental Pharmacology. 2025.
[3] Keizer RJ et al. CPT Pharmacometrics Syst Pharmacol. 2013.
[4] https://www.rdocumentation.org/packages/ggPMX/versions/0.9.2.
[5] Wang W et al. CPT Pharmacometrics Syst Pharmacol. 2016.
[6] Elmokadem A et al. CPT Pharmacometrics Syst Pharmacol. 2019.
Reference: PAGE 34 (2026) Abstr 11918 [www.page-meeting.org/?abstract=11918]
Poster: Methodology - Other topics