Automated, predictive, scalable design and efficient optimization of many-enzyme genetic systems

Engineering large, many-protein genetic systems to maximize their performance remains a Grand Challenge in Metabolic Engineering and Synthetic Biology. Genetic system engineering has three distinct steps: designing the best-possible long DNA sequence for reliable, focused control over its protein expression levels; efficiently optimizing protein expression levels, while constructing and characterizing the fewest genetic system variants; and identifying the most important follow-up protein and strain engineering efforts to improve intrinsic enzyme kinetics and increase precursor metabolic fluxes.

Here, we present an automated pipeline that carries out these steps, combining predictive biophysical models with computational optimization algorithms to: (i) carry out on-demand design of bacterial operons; (ii) efficiently optimize many-enzyme pathway expression levels with the fewest experiments; and (iii) learn the pathway’s intrinsic enzyme kinetic parameters and metabolic precursor fluxes using readily obtainable measurements. We combine biophysical models of transcription, translation, translational coupling, and mRNA stability with 15 design rules for ensuring stable, reliable expression of genome-integrated bacterial operons (Operon Calculator). We apply automated model reduction and genetic algorithm optimization to predict a pathway’s optimal enzyme expression levels, requiring end-product measurements from only ~100 pathway variants (Pathway Map Calculator). We present experimental validation of our pipeline, including a 5-enzyme synthetic Entner-Doudoroff pathway that improves NADPH regeneration by 25-fold in E. coli. Altogether, our pipeline requires only 2 rounds of design-build-test-learn and identifies the most important enzyme/strain bottlenecks, and therefore will dramatically accelerate Metabolic Engineering efforts.