h1

h2

h3

h4

h5
h6
%0 Thesis
%A Osthege, Michael
%T Accelerated bioprocess research by autonomous experimentation and Bayesian modeling
%I RWTH Aachen University
%V Dissertation
%C Aachen
%M RWTH-2024-04287
%P 1 Online-Ressource : Illustrationen
%D 2023
%Z Veröffentlicht auf dem Publikationsserver der RWTH Aachen University 2024
%Z Dissertation, RWTH Aachen University, 2023
%X In times of a fast growing bioeconomy, the fast-paced development of new, and optimization of existing bioprocesses using laboratory automation is one of the most important challenges in bioprocess engineering. In combination with cultivation capabilities at the mL scale, these automation platforms promise to reliably deliver accurate results that would be laborious to obtain by manual labor alone. However, the versatility of bioprocesses and the peculiarities of biochemical procedures have the potential to curb the establishment of automated workflows to the point where they are no longer viable. Consequently, organizations that rely on laboratory automation are in dire need of solutions that enable fast implementation of automation workflows. To accelerate the establishment of such workflows, a process control system and ecosystem of auxiliary software solutions were developed. This new platform architecture and infrastructure enables practicioners to quickly implement cutting-edge autonomous experimentation workflows for their research. The broad applicability of the platform is demonstrated by example projects such as bioprocesses for the production of small molecules or the characterization of whole-cell biocatalysts. To facilitate quantitative data analysis under uncertainty, a bottom-up understanding of calibration models as the linkage between experimental measurement noise and likelihood functions is presented. For multiple application projects, this framework enabled the application of Bayesian hierarchical modeling to datasets obtained on the laboratory automation platform. With examples from classical machine learning to Bayesian optimization with hierarchical differential equation models or Gaussian processes, the application studies demonstrate the effectiveness of combining programmable autonomous experimentation with modern statistical modeling techniques.
%F PUB:(DE-HGF)11
%9 Dissertation / PhD Thesis
%R 10.18154/RWTH-2024-04287
%U https://publications.rwth-aachen.de/record/984915