Machine learning provides researchers a unique opportunity to make metabolic engineering more predictable. In this review, we offer an introduction to this discipline in terms that are relatable to metabolic engineers, as well as providing in-depth illustrative examples leveraging omics data and improving production. We also include practical advice for the practitioner in terms of data management, algorithm libraries, computational resources, and important non-technical issues. A variety of applications ranging from pathway construction and optimization, to genetic editing optimization, cell factory testing, and production scale-up are discussed. Moreover, the promising relationship between machine learning and mechanistic models is thoroughly reviewed. Finally, the future perspectives and most promising directions for this combination of disciplines are examined.
Computational modeling and analysis of metabolic networks has been successful in metabolic engineering of microbial strains for valuable biochemical production. Limitations of currently available computational methods for metabolic engineering are that they are often based on reaction deletions rather than gene deletions and do not consider the regulatory networks that control metabolism. Due to the presence of multi-functional enzymes and isozymes, computational designs based on reaction deletions can sometimes result in strategies that are genetically complicated or infeasible. Additionally, strains might not be able to grow initially due to regulatory restrictions. To overcome these limitations, we have developed a new approach (OptORF) for identifying metabolic engineering strategies based on gene deletion and overexpression.
Predicting cellular responses to perturbations is an important task in systems biology. We report a new approach, RELATCH, which uses flux and gene expression data from a reference state to predict metabolic responses in a genetically or environmentally perturbed state. Using the concept of relative optimality, which considers relative flux changes from a reference state, we hypothesize a relative metabolic flux pattern is maintained from one state to another, and that cells adapt to perturbations using metabolic and regulatory reprogramming to preserve this relative flux pattern. This constraint-based approach will have broad utility where predictions of metabolic responses are needed.
The use of computational models in metabolic engineering has been increasing as more genome-scale metabolic models and computational approaches become available. Various computational approaches have been developed to predict how genetic perturbations affect metabolic behavior at a systems level, and have been successfully used to engineer microbial strains with improved primary or secondary metabolite production. However, identification of metabolic engineering strategies involving a large number of perturbations is currently limited by computational resources due to the size of genome-scale models and the combinatorial nature of the problem. In this study, we present (i) two new bi-level strain design approaches using mixed-integer programming (MIP), and (ii) general solution techniques that improve the performance of MIP-based bi-level approaches. The first approach (SimOptStrain) simultaneously considers gene deletion and non-native reaction addition, while the second approach (BiMOMA) uses minimization of metabolic adjustment to predict knockout behavior in a MIP-based bi-level problem for the first time. Our general MIP solution techniques significantly reduced the CPU times needed to find optimal strategies when applied to an existing strain design approach (OptORF) (e.g., from ∼10 days to ∼5 minutes for metabolic engineering strategies with 4 gene deletions), and identified strategies for producing compounds where previous studies could not (e.g., malate and serine). Additionally, we found novel strategies using SimOptStrain with higher predicted production levels (for succinate and glycerol) than could have been found using an existing approach that considers network additions and deletions in sequential steps rather than simultaneously. Finally, using BiMOMA we found novel strategies involving large numbers of modifications (for pyruvate and glutamate), which sequential search and genetic algorithms were unable to find. The approaches and solution techniques developed here will facilitate the strain design process and extend the scope of its application to metabolic engineering.
Rhodosporidium toruloides is a promising host for the production of bioproducts from lignocellulosic biomass. A key prerequisite for efficient pathway engineering is the availability of robust genetic tools and resources. However, there is a lack of characterized promoters to drive expression of heterologous genes for strain engineering in R. toruloides. This data describes a set of native R. toruloides promoters, characterized over time in four different media commonly used for cultivation of this yeast. The promoter sequences were selected using transcriptional analysis and several of them were found to drive expression bidirectionally. Promoter expression strength was determined by measurement of EGFP and mRuby2 reporters by flow cytometry. A total of 20 constitutive promoters (12 monodirectional and 8 bidirectional) were found, and are expected to be of potential value for genetic engineering of R. toruloides. A set of robust and constitutive promoters to facilitate genetic engineering of R. toruloides is presented here, ranging from a promoter previously used for this purpose (P7, glyceraldehyde 3-phosphate dehydrogenase, GAPDH) to stronger monodirectional (e.g., P15, mitochondrial adenine nucleotide translocator, ANT) and bidirectional (e.g., P9 and P9R, histones H3 and H4, respectively) promoters. We also identified promoters that may be useful for specific applications such as late-stage expression (e.g., P3, voltage-dependent anion channel protein 2, VDAC2). This set of characterized promoters significantly expands the range of engineering tools available for this yeast and can be applied in future metabolic engineering studies.