Stochastic models to optimize biomanufacturing operations

Research output: ThesisPhd Thesis 4 Research NOT TU/e / Graduation NOT TU/e)

Abstract

Biomanufacturing methods use live cells to manufacture vaccines and proteins. The use of live cells introduces several operational challenges, including uncertainty in yield and quality, random batch failures, and challenges in meeting specific production requirements for engineered drugs. In this dissertation, we present optimization models to reduce costs and lead times in biomanufacturing operations. First, we present a stochastic model that balances the risk of batch failures and yield-quality trade-offs to reduce costs in upstream biomanufacturing operations. We develop reliability models for random batch failures, and then provide a Markov decision model to derive the structural properties of the optimal operating policies. Second, we analyze a protein purification problem where each order denotes an engineered protein having specific requirements on the yield and quality. We develop a Markov decision model to optimize the pooling decisions for a fixed sequence of chromatography operations. We partition the state space into distinct decision zones, and then analyze the best starting material and optimal policies that would lead to guaranteed performance outcomes. We present zone-based optimal policies that are easy to implement in practice. Third, we consider the interaction between upstream fermentation and downstream purification operations. We develop a stochastic optimization model to examine the joint decision on chromatography techniques and pooling windows. Then, we optimize the amount of protein to be manufactured in the upstream fermentation operations considering the uncertainty in the yield and quality of the downstream purification operations. This research provides several contributions to theory and practice. First, we provide novel functional relationships between yield, quality and costs in biomanufacturing operations. Next, we develop Markov decision models that capture both biology-level and manufacturing system-level dynamics in a unified framework. We analyze the structural properties, and propose optimal guidelines for industry practices. Our research findings have been developed in close collaboration with the biomanufacturing industry and have been implemented in practice. To facilitate industry implementation, software prototypes have been developed. Applications of operations research are mostly new to both communities. We believe that as more companies embrace operations research, it will be an essential part of the protein research and development processes.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • University of Wisconsin-Madison
Thesis sponsors
Award date31 Jul 2015
Publication statusPublished - 31 Jul 2015
Externally publishedYes

Fingerprint

Dive into the research topics of 'Stochastic models to optimize biomanufacturing operations'. Together they form a unique fingerprint.

Cite this