The metabolome of an organism depends on environmental factors and intracellular regulation and provides information about the physiological conditions. Metabolomics helps to understand disease progression in clinical settings or estimate metabolite overproduction for metabolic engineering. The most popular analytical metabolomics platform is mass spectrometry (MS). However, MS metabolome data analysis is complicated, since metabolites interact nonlinearly, and the data structures themselves are complex. Machine learning methods have become immensely popular for statistical analysis due to the inherent nonlinear data representation and the ability to process large and heterogeneous data rapidly. In this review, we address recent developments in using machine learning for processing MS spectra and show how machine learning generates new biological insights. In particular, supervised machine learning has great potential in metabolomics research because of the ability to supply quantitative predictions. We review here commonly used tools, such as random forest, support vector machines, artificial neural networks, and genetic algorithms. During processing steps, the supervised machine learning methods help peak picking, normalization, and missing data imputation. For knowledge-driven analysis, machine learning contributes to biomarker detection, classification and regression, biochemical pathway identification, and carbon flux determination. Of important relevance is the combination of different omics data to identify the contributions of the various regulatory levels. Our overview of the recent publications also highlights that data quality determines analysis quality, but also adds to the challenge of choosing the right model for the data. Machine learning methods applied to MS-based metabolomics ease data analysis and can support clinical decisions, guide metabolic engineering, and stimulate fundamental biological discoveries.

Over 359 million tons of plastics were produced worldwide in 2018, with significant growth expected in the near future, resulting in the global challenge of end-of-life management. The recent identification of enzymes that degrade plastics previously considered non-biodegradable opens up opportunities to steer the plastic recycling industry into the realm of biotechnology. Here, the sequential conversion of post-consumer polyethylene terephthalate (PET) into two types of bioplastics is presented: a medium chain-length polyhydroxyalkanoate (PHA) and a novel bio-based poly(amide urethane) (bio-PU). PET films are hydrolyzed by a thermostable polyester hydrolase yielding highly pure terephthalate and ethylene glycol. The obtained hydrolysate is used directly as a feedstock for a terephthalate-degrading Pseudomonas umsongensis GO16, also evolved to efficiently metabolize ethylene glycol, to produce PHA. The strain is further modified to secrete hydroxyalkanoyloxy-alkanoates (HAAs), which are used as monomers for the chemo-catalytic synthesis of bio-PU. In short, a novel value-chain for PET upcycling is shown that circumvents the costly purification of PET monomers, adding technological flexibility to the global challenge of end-of-life management of plastics.

Background: Quantification of intracellular metabolite fluxes by ¹³C-tracer experiments is maturing into a routine higher-throughput analysis. The question now arises as to which mutants should be analyzed. Here we identify key experiments in a systems biology approach with a genome-scale model of Saccharomyces cerevisiae metabolism, thereby reducing the workload for experimental network analyses and functional genomics. Results: Genome-scale ¹³C flux analysis revealed that about half of the 745 biochemical reactions were active during growth on glucose, but that alternative pathways exist for only 51 gene-encoded reactions with significant flux. These flexible reactions identified in silico are key targets for experimental flux analysis, and we present the first large-scale metabolic flux data for yeast, covering half of these mutants during growth on glucose. The metabolic lesions were often counteracted by flux rerouting, but knockout of cofactor-dependent reactions, as in the adh1, ald6, cox5A, fum1, mdh1, pda1, and zwf1 mutations, caused flux responses in more distant parts of the network. By integrating computational analyses, flux data, and physiological phenotypes of all mutants in active reactions, we quantified the relative importance of 'genetic buffering' through alternative pathways and network redundancy through duplicate genes for genetic robustness of the network. Conclusions: The apparent dispensability of knockout mutants with metabolic function is explained by gene inactivity under a particular condition in about half of the cases. For the remaining 207 viable mutants of active reactions, network redundancy through duplicate genes was the major (75%) and alternative pathways the minor (25%) molecular mechanism of genetic network robustness in S. cerevisiae.