Skip to main content
Open Access Publications from the University of California

Optimization of microbial cell factories with systems biology

  • Author(s): King, Zachary Andrew
  • Advisor(s): Palsson, Bernhard O
  • et al.

Microbial cell factories can have a transformative impact the chemical industry, but, first, we must meet the challenges of designing and optimizing high-yield cell factory strains. The most popular conceptual model for cell factory optimization is the design-build-test-learn cycle. I present methods that use systems biology to improve the optimization process in each of these steps. First, the build step requires a parts list for a host organism and any heterologous pathways. I present BiGG Models, a database of more than 75 high-quality, manually-curated genome-scale metabolic models that comprise a standardized metabolic parts list. BiGG Models has become the most popular resource in the community for gold-standard genome-scale metabolic models. For the test step, contextualization of omics data is an enormous challenge, and I developed a visualization tool to address this challenge. Escher is a web application for visualizing data on biological pathways. With Escher, users can identify trends in common genomic data types (e.g. RNA-Seq, proteomics, ChIP) and metabolite- and reaction-oriented data types (e.g. metabolomics, fluxomics). For the learn step, genome-scale models can be used to identify general trends in cell factories performance. I introduce a computational method—OptSwap—to predict bioprocessing strain designs by identifying optimal modifications of the cofactor binding specificities of oxidoreductase enzyme and identifying complementary reaction knockouts. I also present an optimization procedure that identifies optimal cofactor- specificity “swaps” for improving theoretical yield in genome-scale metabolic models. Swapping the cofactor specificity of central metabolic enzymes is shown to increase NADPH production and increase theoretical yields for many native and non-native products. Last, the design step requires models that can successfully predict phenotype from genotype. I assess the predictive capabilities of existing models of E. coli through literature mining. I simulate strains from the literature in six historical genome-scale models of E. coli and report that the predictive power of the models has increased as they have expanded in size and scope. Together, these studies provide a path toward successfully applying systems biology methods to optimizing microbial cell factories.

Main Content
Current View