In August 2018, the nlmixr team announced the first baby step by nlmixr – the release of nlmixr 1.0. Today, we proudly announce the arrival of nlmixr’s adulthood: nlmixr 2.0 is here. It has taken a bit longer than we’d like to see nlmixr getting through its adolescence, but that day has arrived! Lots of hard work has gone into the new version over the past 2.5 years, and we think the pharmacometrics community will be very happy to see all the new functionality & features in nlmixr 2.0.
A few highlights include:
• Automatic covariate search with covarSearchAuto
• Boostrapping with bootstrapFit
• Time-varying covariates in SAEM
• Parallel SAEM
nlmixr 2.0 no longer needs Python for its symbolic algebra computations (e.g., calculating derivatives for FOCEi types of algorithms). This change not only makes nlmixr faster, it also leaves a smoother user experience and simplifies installation and maintenance.
For the geekier modelers among you, you’ll find a few more advanced tricks have been baked in:
• ETA support on dose-based parameters like lag time and infusion duration for FOCEi
• Support of lag/diff and NONMEM-style NEWIND flag
• More logical expression support including ifelse() and logical expressions (for example, KA = (FED==1)*KA1 + (FED != 1)*KA2)
• Mean model support (i.e. no ETAs) for FOCEi
• More types of residual model support in both SAEM and FOCEi
• Censoring support using CENS and LIMIT columns
As always, the nlmixr team are grateful for the heart-warming support we’ve received from our users. You are our strength and inspiration. Our goal here is to make nlmixr a PKPD modeling tool that is truly of the modelers, by the modelers and for the modelers.
We hope you’ll enjoy this youthful, vibrant yet mature nlmixr. Please continue to send us your ideas, suggestions and bug reports via our GitHub page.