Roadmap

This page lists planned and in-progress work for ferx. Timelines are indicative and may shift as priorities evolve. Feel free to reach out with any feature requests, research project proposals, or feedback.

Short term (Q2–Q3 ’26)

• Improve testing infrastructure with larger coverage, with more real datasets and models.
• Support for TTE / survival modeling and joint PK/PD–survival modeling.
• Support for stiff ODE solver(s).
• NONMEM and nlmixr2 model import. nlmixr2 models can be supported in a non-lossy way, since they can be read from the R object.
• Increase modularity of ferx, so external users can more easily write extensions with minimal (if any) modifications to ferx core code.
• Support for FREM estimation (including conversion to FFEM models).
• Support for covariate search approaches (GAM, SCM).
• Exact analytic sensitivities through the ODE solver (ODE models currently use finite differences).
• Support for use of priors (like in NONMEM with penalties added to the likelihood, but with easier implementation).
• Support for mixture models.

Medium term (Q4 ’26)

• Release of v1.0.0.
• Benchmark vignette and subsequent publication in a journal.
• Support for count data models (Poisson).
• Support for Item Response Theory models.
• Support for optimal design analyses. At least computation of the FIM, and basic optimal design searches. Potentially in a separate R package.

Ongoing research projects

• Copula-estimation of between-subject variability with SAEM.
• Improved support for neural nets: deep compartmental models and neural ODEs.