Application of genetic algorithms and response surface analysis for the optimization of batch chromatographic systems

Abstract

A trend to scientifically sound, robust and especially accelerated downstream process development in the biopharmaceutical industry is encouraged by the FDA [1,2]. The underlying methodology – high throughput process development – is an interplay of high throughput experimentation, usage of adequate analytics and model based experimentation or process development. In this study, high-throughput batch chromatographic methods were combined with design of experiments, genetic algorithm and response surface analysis to optimize a cation exchange step. The optimization was successful in batch mode and validated in packed bed column mode. Full automation was achieved by establishing a method to automate pH and salt concentration adjustment on liquid handling stations. Several process optima were identified by the genetic algorithm. It was demonstrated that the initial population design influenced the number of optima found during a genetic algorithm optimization procedure. The mere application of response surface analysis on the experimental results showed that for systems with several optima no distinct statements on parameter dependency are achieved. A combination of genetic algorithm, design of experiments and response surface analysis showed to be the most efficient data usage during process optimization if no information on the process landscape investigated is available.