Introducing nlmixr

There’s a distinct shortage of tools for performing effective pharmacometric data analysis. There’s NONMEM, the gold standard; there’s Monolix; and there’s Phoenix WinNonlin. All of these tools do a good enough job, but are united by two critical issues: first, they’re closed and proprietary, and second, they cost the equivalent of a small country’s budget to run each year.

There’s a crying need for a tool that can do the same things just as well as the big dogs can, but won’t cripple small CROs and less well-funded academic institutions with costs. Starting with the NLME Consortium, there have been several attempts to this in the past – Monolix was one, but after a strong start as an open-source, free tool, it went commercial. The NLME Consortium itself morphed into DDMoRe, which gave us a model library, standards and Thoughtflow, but no new tool.

With the current renaissance of free and open-source tools in the field of pharmacometrics and quantitative systems pharmacology, and robust, open, platforms like R, Julia and Stan to build on, we believe there has never been a better opportunity to build one.

In case you haven’t come across it, nlmixr is an R package for fitting general dynamic models, pharmacokinetic (PK) models and pharmacokinetic-pharmacodynamic (PKPD) models in particular, with either individual data or population data. Although we’re very pleased with progress, nlmixr is still experimental – you can find out more and download the latest release here. Wenping Wang (@wenping_wang) is our lead developer, supported by Matt Fidler, Yuan Xiong (@yxiong), Rik Schoemaker, and Justin Wilkins (@JustinWilkins).

So far, we’ve integrated modules for:

  • nonlinear dynamic models of individual data;
  • one to three linear compartment models of population data with first order absorption, or i.v. bolus, or i.v. infusion using R’s nlme algorithm;
  • general dynamic models defined by ordinary differential equations (ODEs) of population data using R’s nlme algorithm;
  • general dynamic models defined by ordinary differential equations (ODEs) of population data by the Stochastic Approximation Expectation-Maximization (SAEM) algorithm; and
  • generalized non-linear mixed-models (possibly defined by ODEs) of population data by the adaptive Gaussian quadrature algorithm.

We’re hard at work expanding nlmixr’s functionality, and hope to release regular updates over the coming months. FOCE with interaction has been prototype and is in testing… but more on that later.

We’re going to post here occasionally, to keep everyone informed of what we’re up to, and to get feedback from you, the community. Stick around for the ride. It’s going to be fun.

(Oh, and you can follow us on Twitter at @nlmixr.)



1 Comment

  1. Marc Lavielle | | Reply

    for your information, Monolix is (and will remain) totally free for academia

Leave a Reply

Your email address will not be published. Required fields are marked *