Lars Freier

Framework for Kriging-based iterative experimental analysis and design: Optimization of secretory protein production in Corynebacterium glutamicum

The production of bulk enzymes used in food industry or organic chemistry constitutes an important part of industrial biotechnology. The development of production processes for novel proteins comprises a variety of biological engineering and bioprocess reaction engineering factors. The combinatorial explosion of these factors can be effectively countered by combining high‐throughput experimentation with advanced algorithms for data analysis and experimental design. We present an experimental optimization strategy that merges three different techniques: (1) advanced microbioreactor systems, (2) lab automation, and (3) Kriging‐based experimental analysis and design. This strategy is demonstrated by maximizing product titer of secreted green fluorescent protein (GFP), synthesized by Corynebacterium glutamicum, through systematic variation of CgXII minimal medium composition. First, relevant design parameters are identified in an initial fractional factorial screening experiment. Then, the functional relationship between selected media components and protein titer is investigated more detailed in an iterative procedure. In each iteration, Kriging interpolations are used for formulating hypotheses and planning the next round of experiments. For the optimized medium composition, GFP product titer was more than doubled. Hence, Kriging‐based experimental analysis and design has been proven to be a powerful tool for efficient process optimization.

Multi‐Objective Global Optimization (MOGO): Algorithm and Case Study in Gradient Elution Chromatography

Biotechnological separation processes are routinely designed and optimized using parallel high-throughput experiments and/or serial experiments. Well-characterized processes can further be optimized using mechanistic models. In all these cases - serial/parallel experiments and modeling - iterative strategies are customarily applied for planning novel experiments/simulations based on the previously acquired knowledge. Process optimization is typically complicated by conflicting design targets, such as productivity and yield. We address these issues by introducing a novel algorithm that combines recently developed approaches for utilizing statistical regression models in multi-objective optimization. The proposed algorithm is demonstrated by simultaneous optimization of elution gradient and pooling strategy for chromatographic separation of a three-component system with respect to purity, yield, and processing time. Gaussian Process Regression Models (GPM) are used for estimating functional relationships between design variables (gradient, pooling) and performance indicators (purity, yield, time). The Pareto front is iteratively approximated by planning new experiments such as to maximize the Expected Hypervolume Improvement (EHVI) as determined from the GPM by Markov Chain Monte Carlo (MCMC) sampling. A comprehensive Monte-Carlo study with in-silico data illustrates efficiency, effectiveness and robustness of the presented Multi-Objective Global Optimization (MOGO) algorithm in determining best compromises between conflicting objectives with comparably very low experimental effort.

Robust Multi-Objective Global Optimization of Stochastic Processes with a Case Study in Gradient Elution Chromatography†

A novel algorithm for robust multi‐objective process optimization under stochastic variability of environmental variables is introduced and applied to a case study in gradient elution chromatography. Process variability is accounted for by simultaneously optimizing several scenarios with random but fixed values of the environmental variables. These iterative optimizations are synchronized by planning the same experiments for all scenarios. Experiments are designed by maximizing the cumulative expected hypervolume improvement as predicted by several Gaussian process regression models. A straightforward method is presented for estimating the expected Pareto front and its variability based on the resulting data that maintains traceability of the corresponding process parameters. This information is required for robust process optimization, that is, determination of Pareto optimal processes that fulfil specific minimal criteria with a certain confidence. The presented algorithm can generally be applied to both in silico and wet lab experiments but involves an increased experimental effort as compared to the deterministic case.

Generic Protocol for Optimization of Heterologous Protein Production Using Automated Microbioreactor Technology

A core business in industrial biotechnology using microbial production cell factories is the iterative process of strain engineering and optimization of bioprocess conditions. One important aspect is the improvement of cultivation medium to provide an optimal environment for microbial formation of the product of interest. It is well accepted that the media composition can dramatically influence overall bioprocess performance. Nutrition medium optimization is known to improve recombinant protein production with microbial systems and thus, this is a rewarding step in bioprocess development. However, very often standard media recipes are taken from literature, since tailor-made design of the cultivation medium is a tedious task that demands microbioreactor technology for sufficient cultivation throughput, fast product analytics, as well as support by lab robotics to enable reliability in liquid handling steps. Furthermore, advanced mathematical methods are required for rationally analyzing measurement data and efficiently designing parallel experiments such as to achieve optimal information content. The generic nature of the presented protocol allows for easy adaption to different lab equipment, other expression hosts, and target proteins of interest, as well as further bioprocess parameters. Moreover, other optimization objectives like protein production rate, specific yield, or product quality can be chosen to fit the scope of other optimization studies. The applied Kriging Toolbox (KriKit) is a general tool for Design of Experiments (DOE) that contributes to improved holistic bioprocess optimization. It also supports multi-objective optimization which can be important in optimizing both upstream and downstream processes.

Kriging with trend functions nonlinear in their parameters: Theory and application in enzyme kinetics

Kriging is an interpolation method commonly applied in empirical modeling for approximating functional relationships between impact factors and system response. The interpolation is based on a statistical analysis of given data and can optionally include a priori defined trend functions. However, Kriging can so far only be used with trend functions that are linear with respect to the parameters. In this contribution, we present an extension of the Kriging approach for handling trend functions that are nonlinear in their parameters. Our approach is based on Taylor linearization combined with an iterative parameter estimation procedure whose solution is practically computed via a root finding problem. We demonstrate our novel approach with measurement data from the application field of biocatalysis.