A versatile active learning workflow for optimization of genetic and metabolic networks

Abstract

Optimization of biological networks is often limited by wet lab labor and cost, and the lack of convenient computational tools. Here, we describe METIS, a versatile active machine learning workflow with a simple online interface for the data-driven optimization of biological targets with minimal experiments. We demonstrate our workflow for various applications, including cell-free transcription and translation, genetic circuits, and a 27-variable synthetic CO2-fixation cycle (CETCH cycle), improving these systems between one and two orders of magnitude. For the CETCH cycle, we explore 1025 conditions with only 1,000 experiments to yield the most efficient CO2-fixation cascade described to date. Beyond optimization, our workflow also quantifies the relative importance of individual factors to the performance of a system identifying unknown interactions and bottlenecks. Overall, our workflow opens the way for convenient optimization and prototyping of genetic and metabolic networks with customizable adjustments according to user experience, experimental setup, and laboratory facilities.

Enzymatic Conversion of CO2: From Natural to Artificial Utilization

“…This is largely because the individual parts used to reconstruct these systems are derived from drastically different biological backgrounds, which makes their interactions hard to predict. Recent efforts have therefore implemented a machine learning-guided workflow called METIS to explore the vast combinatorial space of reaction conditions. By screening the efficiency of only ∼1000 different pathway variants over a total of eight rounds of active learning, the productivity of the CETCH cycle could be improved roughly 10-fold compared to the previously best pathway combination. , In theory, workflows like METIS can be used from the get-go to identify and optimize reaction conditions for efficient realization of CO 2 fixation cascades.…”

“…An efficient way to exert an external driving force onto an enzymatic reaction system is the removal of the formed product by an additional (irreversible) reaction. Owing to the fact that biocatalysts usually require similar reaction conditions, the addition of a subsequent enzymatic step to a carboxylation reaction is often straightforward and mirrors nature’s CO 2 fixation strategies, where carboxylation reactions are part of biosynthetic pathways or cycles. ,,,,− The efficiency of a reaction can be evaluated from a thermodynamic point of view by estimating the Gibbs free energy demand of the individual reactions. Dedicated tools allow to calculate the Δ G of small cascades and even entire pathways at physiological or process conditions and therefore to evaluate their thermodynamic feasibility …”

Open-endedness in synthetic biology: a route to continual innovation for biological design

“…Finally, Bayesian Optimization [134] has recently gained popularity in synthetic biology [37,120,85,114] and might offer a possible route to open-endedness. A machine learning surrogate model, such as a Gaussian Process or Random Forest, is used to grade the entities based on a so-called acquisition function.…”

Engineering the Substrate Specificity of Toluene Degrading Enzyme XylM Using Biosensor XylS and Machine Learning

“…Therefore, a transcriptional regulator that recognizes substrates or products of a target enzyme is useful for high-throughput and sensitive evaluation or selection of desired enzyme variant activities. Many researchers have used such screening systems to modify hosts or enzymes to improve the yield of desired products. − In addition, the screening systems can be used to modify enzymes that are difficult to analyze in vitro, such as membrane proteins and those catalyzing multistep reactions.…”