Multi-objective optimization in population pharmacokinetic model selection and optimization: application of NSGA-II...

Area of Science:

• Pharmacometrics
• Computational Biology
• Systems Pharmacology

Background:

• Model selection often relies on a mix of objective and subjective criteria.
• Quantifying all desirable model characteristics numerically can be challenging.
• Multi-objective optimization (MOO) offers a framework to balance competing objectives in model selection.

Purpose of the Study:

• To apply MOO using the non-dominated sorting genetic algorithm II (NSGA-II) for population pharmacokinetic model selection.
• To simultaneously optimize objective function value and the number of estimated parameters.
• To compare MOO with traditional stepwise search and single-objective hybrid genetic algorithms (SOHGA).

Main Methods:

• Population pharmacokinetic models were built for 17-DMAG, quetiapine, clozapine, and ziprasidone.
• Non-dominated sorting genetic algorithm II (NSGA-II) was used for multi-objective optimization.
• Comparisons were made with stepwise search, SOHGA, and local downhill search.

Main Results:

• A trade-off was observed: lower objective function values generally corresponded to more estimated parameters.
• The number of non-dominated solutions (Pareto front) varied for each drug: 17 for DMAG, 9 for ziprasidone, 9 for clozapine, and 13 for quetiapine.
• SOHGA-selected optimal models were found on the Pareto front for DMAG, ziprasidone, and clozapine.

Conclusions:

• MOO provides objective transparency regarding the cost of tradeoffs between competing model objectives.
• This approach aids researchers in contextualizing subjective criteria like biological plausibility for model selection.
• MOO facilitates better alignment of model selection with clinical context in pharmacokinetic studies.