The Pathway Map Calculator: an Automated Learning Algorithm to Determine a Pathway's Optimal Expression Levels

Heterologous metabolic pathways enable microbes to manufacture high-value chemicals from low-value feedstock. However, tuning the pathway and network expression levels to maximize the organism's productivity remains a major challenge, particularly when many enzymes, transporters, cofactors, and auxiliary proteins are involved.  The current industry trend is to perform high-throughput combinatorial DNA assembly, mutagenesis, and characterization to identify the best organism variants; however, because the search space is so large, new algorithms are needed to extract actionable knowledge from such large data-sets. Here, we present the Pathway Map Calculator, an automated learning algorithm that predicts a pathway/network's optimal expression levels when given the results of a combinatorial promoter or RBS assembly with end-product measurements. Using experimentally validated examples, we show how the Pathway Map Calculator is used to extract a pathway's enzymatic kinetic parameters from sparse, multi-dimensional data-sets, followed by predicting the promoter/RBS sequences that optimally balance the pathway and achieve maximum productivity. Our pathway examples include a 3-enzyme terpenoid biosynthesis pathway (Farasat et. al., Mol. Sys. Bio., 2014) and a 5-enzyme Entner-Doudoroff pathway for rapid NADPH regeneration (Ng. et. al., Metabolic Engineering, 2015). Using in silico examples, we show that the algorithm scales well when simultaneously modifying many protein expression levels, which is now possible via Cas9-based genome mutagenesis.