E. Niclas Jonsson, Lars Lindbom, Jakob Ribbing and Mats O. Karlsson
Department of Pharmacy, Uppsala University, Sweden
From a data analyst point of view, the majority of the time in a population analysis using NONMEM, is spent on “house-keeping tasks” like data file preparation and run management. Examples of the latter includes tabulation of parameter estimates and termination codes from multiple runs, serial runs in covariate model buidling and various validation procedures. If done manually, these tasks are not only time consuming and inefficient, but also a clear source for user errors. On the other hand, due to their repetitive nature, they can often be implemeted in scripts or small programs, which quickly and reproducibly accomplish the same thing.
perl is a free, multi-platform scripting language that was originally designed for handling text files. Today, perl is a full fetured programming language, with the main difference from “real” programming languages, like C being that it is interpreted rather than compiled. Since NONMEM is used on many different computer platforms and uses text files for all its input and output, perl is the ideal language for the kind of programs outlined above.
Perl-speaks-NONMEM (PsN) is a support library for NONMEM related programming. It includes low level, task oriented, functions that can be used as building blocks for specialized scripts. Examples of such functions are parameter estimate extraction from output files, data file subsetting, setting of initial estimates from a previous run and running of NONMEM with a re-try option with perturbed initial estimates.
A couple of scripts are distributed together with PsN, for example a stepwise covariate model building procedure and scripts for jacknife, bootstrap and cross-validation.
PsN is free of charge and can be obtained from http://www.biof.uu.se.
Reference: PAGE 9 () Abstr 124 [www.page-meeting.org/?abstract=124]
Poster: poster