(500a) The Automated Design of Multi-Enzyme Pathways for Efficient and Scalable Optimization in High-Dimensional Expression Spaces

The optimization of heterologous metabolic pathways for chemical production requires the precise tuning of enzyme expression levels to achieve balanced enzyme kinetics and maximum pathway productivity. The state-of-the-art strategy to achieve this goal is to test libraries of genetic parts, such as promoters and ribosome binding sites (RBS), which coarsely cover the high-dimensional enzyme expression space of a given pathway. As pathway size increases, so does the number of genetic manipulations needed to screen for an optimal pathway variant, reducing the experimental viability of library-based approaches on large pathways. To overcome this combinatorial hurdle, we have developed the Pathway Map Calculator, an automated algorithm that utilizes a small number of characterized pathway variants to learn the expression-flux relationship in a high-dimensional space, in order to accurately predict the optimal enzyme expression levels that maximize the pathwayâ??s flux. We confirm the algorithmâ??s accurate prediction optimal enzyme expression levels for pathways of varying size and complexity using in silico pathway examples, and have successfully optimized two heterologous pathways in E. coli with this method: the neurosporene producing crtEBI pathway of R. sphaeroides and the NADPH regenerating Entner-Doudoroff pathway of Z. mobilis. The Pathway Map Calculator is the final piece in a larger tripartite computational pipeline: the Operon Calculator arranges and sequence optimizes a pathway for a given host, the RBS Library Calculator develops optimal RBS libraries for tuning each enzymeâ??s expression levels, and the Pathway Map Calculator determines the best enzyme expression levels to use for each enzyme to maximize productivity. This pipeline enables the sequence level design, testing, and optimization of a pathway for maximal production of a desired metabolite in any desired prokaryotic host.