Dear all, In the last couple of months I have been working on NMGUI2, a graphical user interface for pharmacometric workflows with NONMEM, R, RStudio and PSN, mainly for my own use, but would like to share it. It runs in Python and should therefore natively run on Windows, MacOS and LINUX as long as Python (version >3.10) is installed.
It is free and open-source and can be downloaded through https://github.com/robterheine/nmgui2 . Contributions or bug reports are warmly welcomed! The current version of NMGUI (2.5) has the following features: Model browser Opens a directory and displays all NONMEM control stream files (.mod/.ctl) in a sortable, filterable table. Columns include OFV, ΔOFV relative to a reference model, minimisation status, covariance step result, condition number, estimation method, number of individuals and observations, number of estimated parameters, AIC, and runtime. Rows are colour-coded: green for converged, red for failed, orange for stale (control stream or dataset modified after the last run). Free-text search and filter buttons (All / Completed / Failed / Starred) narrow the list quickly. Model comparison Right-click any model to compare it with another. A side-by-side dialog shows parameter estimates from both models with aligned rows, and a statistics strip with ΔOFV, ΔAIC, ΔBIC, difference in parameter count, LRT p-value, and a plain-English verdict (nested/non-nested, significant or not). Model workbench A sortable table of all completed models in the current directory with OFV, ΔOFV, AIC, BIC, and LRT p-values computed against a user-selected reference. Useful at the end of a covariate search for a quick ranked overview. Exportable to CSV. Model editor Syntax-highlighted editor for control streams with a Save button. Model ancestry tree An interactive node graph of model genealogy, based on the standard PsN parent-model header line or a manually set parent. Zoom, pan, double-click a node to select that model in the table. Useful as a visual record of the model-building sequence. Running models Any PsN tool can be launched directly: execute, vpc, bootstrap, scm, sir, cdd, npc, sse. Each run opens in its own floating window with a live console, an iteration and OFV progress indicator (updated in real time from NONMEM output), an elapsed timer, and a Stop button offering either a gentle stop (allows PsN to finish writing output) or an immediate force kill. Multiple models can run simultaneously. An Active & Recent Runs table in the Run panel shows all live and historical runs for the current project folder, persists across restarts, and lets you click a live row to raise its window. Run records Every run automatically creates an immutable audit record containing a unique run ID, SHA-256 hash of the control stream and dataset at the time of the run, NONMEM/PsN/NMGUI versions, start and end timestamps, duration, final OFV, minimisation status, covariance status, and the full command used. Output viewer The .lst file is rendered as structured HTML inside the app: a summary strip with key results, an estimation steps table for chained $EST runs (e.g. FO → FOCE → IMP) showing per-step OFV, ΔOFV, runtime, significant digits and termination status, NM-TRAN compilation warnings, convergence table, parameter estimates with standard errors, ETABAR with p-values, shrinkage (ETA and EPS), correlation and covariance matrices, and eigenvalues with condition number. The full HTML can be exported and opened in a browser. Parameter table Full THETA/OMEGA/SIGMA table with parameter names (parsed from comments in the control stream), estimates, standard errors, RSE%, 95% confidence intervals, and standard deviation for variance parameters. BLOCK(n) and BLOCK(n) SAME designs are handled correctly. Collapsible by section. Exportable to CSV and as a self-contained HTML report. Goodness-of-fit diagnostics Native diagnostic plots that do not require R: DV vs PRED, DV vs IPRED, CWRES vs TIME, CWRES vs PRED, CWRES histogram with normal distribution overlay, QQ plot with Shapiro-Wilk test and 95% confidence band, and ETA vs covariate scatter plots with LOESS overlay. Auto-loads the sdtab file when a model is selected. Individual fits Paginated DV/IPRED/PRED vs TIME plots per subject, with a configurable number of panels per page. OFV waterfall Ranked ΔOFV bar chart across all completed models in the current directory — useful as a final overview of the covariate search results. Data explorer Interactive scatter plot of any table file with any X/Y combination, colour by any column, multi-condition filtering, and a paginated data table view. VPC Visual Predictive Checks generated via three R backends: vpc, xpose, or xpose4. Options include prediction-corrected VPC, stratification by any column (with validation), configurable prediction and confidence intervals, LLOQ shading, log Y axis, and number of bins. The underlying R script is editable before running. Output is displayed as a PNG inside the app and can be saved as a high-resolution PNG or vector PDF. Parameter uncertainty assessment Run a new bootstrap or SIR via PsN or load an existing results folder. Produces a summary and overall assessment report. QC report Right-click any completed model to generate a self-contained HTML quality-control report with a PASS/WARN/FAIL checklist covering: termination status, covariance step, condition number, maximum %RSE, parameter correlations, ETA and EPS shrinkage, ETABAR p-values, and OMEGA boundary proximity. Dataset integrity checks At scan time, each model's data file is automatically checked for: missing file, column-count mismatches between header and data, non-monotonic TIME within an individual, duplicate dose records, extreme DV values, and high BLQ proportion. Issues appear in the Info panel. Model annotations Each model can be tagged (Base / Candidate / Final / Reject), given a one-line comment visible in the table, and given free-form notes. All annotations are persisted in a project metadata file and survive rescans. History tab Chronological log of all runs started from NMGUI2 (last 200 entries across all projects), with status, model name, tool, full command, start time and duration. Settings Dark and light theme (follows the system or manually overridden). Path configuration for NONMEM, PsN, R and RStudio. Directory bookmarks for quick access to frequently-used project folders. Optional GitHub update check. Cheers, Rob De informatie in dit bericht is uitsluitend bestemd voor de geadresseerde. Aan dit bericht en de bijlagen kunnen geen rechten worden ontleend. Heeft u deze e-mail onbedoeld ontvangen? Dan verzoeken wij u het te vernietigen en de afzender te informeren. Openbaar maken, kopieren en verspreiden van deze e-mail of informatie uit deze e-mail is alleen toegestaan met voorafgaande schriftelijke toestemming van de afzender. Het Radboudumc staat geregistreerd bij de Kamer van Koophandel in het handelsregister onder nummer 80262783. The content of this message is intended solely for the addressee. No rights can be derived from this message or its attachments. If you are not the intended recipient, we kindly request you to delete the message and inform the sender. It is strictly prohibited to disclose, copy or distribute this email or the information inside it, without a written consent from the sender. Radboud university medical center is registered with the Dutch Chamber of Commerce trade register with number 80262783.
