Design of Class Structures and Inter Class Communication Interface : Computer Methods used in Non-linear Mixed Effects Modeling

Abstract: The field of pharmacokinetics and pharmacodynamics describes the interactionbetween drugs and the human body from the time the drug is absorbed in the blood stream, distributed in the body, metabolized and finally excreted. The behavior can be described with nonlinear mixed effects models. NONMEM is a software for estimating parameters of such models and simulation of data. This thesis describes the design of Perl-speaks-NONMEM (PsN) which is a software framework for control of different aspects of NONMEM such as creating input, performing execution and output interpretation and presentation. The framework aims to simplify implementation of a range of statistical tools not featured in NONMEM. PsN includes the Bootstrap for measuring statistical accuracy and precision, the Case Deletion Diagnostics (CDD) for identifying influential individuals in medical studies and the Log Likelihood Profiling (LLP) for estimating asymmetrical confidence intervals for parameter estimates. PsN interfaces with various GRID systems: Sun Grid Engine, Plattform LSF and openMosix to enable parallel execution of NONMEM instances. The implementation of PsN was aided with custom built tools for generating class definitions, constructors and from UML diagrams. PsN also provides a set of command line interface tools that simplify usage for researchers without programming experience.