A flexible and transparent MATLAB framework for empirical and mechanistic pharmacometric modelling
Niklas Hartung, Wilhelm Huisinga
Institute of Mathematics, University of Potsdam, Germany
Objectives: There exist a number of established software tools for well-defined modelling purposes, like NONMEM or Monolix [1-2] for the population analysis of clinical data based on classical compartment models; or GastroPlus, PK-Sim and SimCYP [3-5] to study absorption, special subpopulations or drug-drug interactions. Many questions, however, require the development of new models that go beyond a given, hard-coded model, often starting with a simple model that is subsequently refined to account for the most relevant processes of the question at hand. None of the established software tools fully supports this modelling process in terms of the desirable transparency, flexibility and support for handling large data bases of drug and species-specific data. The objective was to develop a MATLAB-based modelling toolbox to fill this gap.
Methods: Building on a MATLAB toolbox that has been used over the years in teaching physiologically-based pharmacokinetic modelling during the A2-module of the PharMetrX PhD program [6] and on previous work on lumping [7] and pharmacokinetics of monoclonal antibodies [8-10], we developed a novel modelling environment written in the MATLAB language. Core values guiding the toolbox development were flexibility and transparency. The central infrastructure was implemented using an object-oriented approach. The toolbox offers support to handle physiological and drug databases, checks unit compatibility during any computation, and allows to document and track assumptions underlying experimental data, thereon based derived parameter values or modelling assumptions. All databases are customizable and there are no restrictions on the models that can be implemented.
Results: The modelling framework was developed with MATLAB (R2018b). To support computation with units during a complete modelling workflow, we extended the contributed MATLAB Physical Units toolbox (version 4.1.0.0) [11]. Any quantity computed during modelling exercise comes with an associated unit, and unit consistency of operations is enforced. A physiological database and a compound database are provided, as well as several scaling methods for the creation of virtual populations and prediction of mixed drug-species parameters like tissue partition coefficients. All databases, scaling and prediction methods are annotated with metadata (e.g., species origin or scalability of a parameter) that are propagated through model development, making the implications of these assumptions during modelling transparent. A variety of models, ranging from empirical and lumped mechanistic to mechanistic PBPK models for small molecules and monoclonal antibodies, are implemented. A modular specification of model components allows high-level modelling at the scripting level, as illustrated by prepared demo projects. To allow most wide-spread use, the developed MATLAB modelling toolbox is contributed as open-source to the community.
Conclusions: The developed MATLAB toolbox provides a flexible and transparent pharmacometric modelling environment with unit and assumption tracking capacities. It can be used for projects requiring capabilities to develop new models or customize physiological and drug-related parameters and prediction/extrapolation methods.
References:
[1] Beal S et al. 1989-2018. NONMEM User’s Guides. Icon Development Solutions, Ellicott City, MD, USA.
[2] http://lixoft.com/products/monolix/
[3] https://www.simulations-plus.com/software/gastroplus/
[4] https://www.open-systems-pharmacology.org/
[5] https://www.certara.com/software/pbpk-modeling-and-simulation/
[6] https://www.pharmetrx.de/
[7] Huisinga W et al. CPT:PSP (2012) 1:e4
[8] Fronton L et al. J PKPD (2014) 41(2):87–107
[9] Huisinga et al. (2016) “Target-driven pharmacokinetics of biotherapeutics” in: Zhou, Theil (Eds.), Wiley
[10] Fuhrmann S et al. J PKPD (2017) 44(4):351–374
[11] Physical Units toolbox, version 4.1.0.0 https://de.mathworks.com/matlabcentral/fileexchange/38977-physical-units-toolbox
