Microbial Strain Design for Biochemical Production Using Mixed-integer Programming Techniques

Metabolic engineering aims to improve biochemical production in a biological system using genetic modifications. To identify these modifications, conventional approaches have focused on local metabolic pathways without considering the global effects on metabolic behavior. Computational approaches using genome-scale models have successfully identified genetic modifications while considering their global effects on metabolism. However, these approaches can take a significant amount of time to find the optimal modifications due to the large size of metabolic networks. In this work, we present mixed-integer programming (MIP) methods and analysis of bounds on dual variables for fast and effective identification of genetic modification strategies. Using the bi-level optimization approach by Kim and Reed, 2010, we demonstrate how our MIP methods significantly improve performance of strain design algorithms.