Woojin Jung 1, Soyoung Lee 2, Jung-woo Chae 2,3,4, Hwi-yeol Yun 2,3,4
1 College of Pharmacy, CHA University (Pocheon, Republic of Korea), 2 College of Pharmacy, Chungnam National University (Daejeon, Republic of Korea), 3 Senior Health Convergence Research Center, Chungnam National University (Daejeon, ), 4 Department of Bio-AI Convergence, Chungnam National University (Daejeon, Republic of Korea)
Objectives:
NONMEM remains the gold standard for population pharmacokinetic/pharmacodynamic (PK/PD) modeling and simulation. Despite its analytical robustness, routine modeling workflows often rely on command-line operations and manual file management, which can hinder efficiency, reproducibility, and accessibility, particularly in multi-run or iterative development settings. To address these limitations, we developed nmbench, a lightweight and extensible extension for Visual Studio Code (VS Code) that functions as an integrated workbench for NONMEM. nmbench is designed to centralize model development, execution, and diagnostics within a modern development environment, while maintaining compatibility with established tools such as PsN (Perl-speaks-NONMEM) and R-based diagnostic workflows.
Methods:
nmbench was implemented in TypeScript using the VS Code extension API and structured around a workflow-oriented architecture. The extension provides a hierarchical, tree-based visualization of NONMEM project directories, enabling structured management of control streams, output files, and run histories. Visualization of run status, objective function value (OFV) changes, and estimation summaries is supported through dynamic parsing of .mod, .ctl, and .lst files.
Core PsN utilities, including execute, vpc, and bootstrap were integrated into the graphical interface, allowing users to initiate and manage modeling tasks without leaving the development environment. Editor-level shortcut buttons facilitate rapid model execution and data inspection directly from control files. The system is designed to support interoperable workflows involving NONMEM, PsN, and R scripting, thereby reducing fragmentation between analytical tools.
Results:
nmbench provides an integrated suite of diagnostic and visualization tools aimed at improving model traceability and iterative refinement. The built-in .lst viewer enables structured and searchable inspection of NONMEM output sections, including estimation summaries and covariance diagnostics. An interactive, Plotly-based matrix viewer supports cross-run comparison of parameter estimates and objective function changes, facilitating transparent evaluation of model development trajectories. A dedicated table viewer enables visualization of output tables through histograms and time-series plots, supporting rapid exploratory diagnostics.
In addition, nmbench allows execution of user-defined R scripts within the project context, with default templates for generating goodness-of-fit (GOF) plots, visual predictive checks (VPC), and individual diagnostic plots using the xpose and xpose4 packages. By consolidating execution, inspection, and visualization within a single environment, nmbench minimizes context switching between command-line tools, external plotting software, and file systems. The extension is optimized for NONMEM version 7.5.1 and PsN version 5.5.0, and is compatible with Windows and macOS systems supported by VS Code, including remote and container-based development environments.
Conclusions:
nmbench provides a streamlined, reproducible, and user-friendly modeling environment by combining the robust analytical capabilities of PsN with the modern interface of Visual Studio Code. As a lightweight VS Code extension, nmbench offers high flexibility and ease of deployment across platforms, including remote computing environments, making it especially suitable for distributed research workflows and remote-workspace-based modeling tasks. Future developments will include integrated diagnostics, and enhanced visualization modules to support model-informed drug development (MIDD)
References:
[1] Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput Methods Programs Biomed. 2005 Sep;79(3):241-57.
[2] Lindbom L, Ribbing J, Jonsson EN. Perl-speaks-NONMEM (PsN)–a Perl module for NONMEM related programming. Comput Methods Programs Biomed. 2004 Aug;75(2):85-94.
[3] https://github.com/tnzo12/nmbench
Reference: PAGE 34 (2026) Abstr 11933 [www.page-meeting.org/?abstract=11933]
Poster: Software Demonstration