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.

Fatty alcohols are widely used in various applications within a diverse set of industries, such as the soap and detergent industry, the personal care, and cosmetics industry, as well as the food industry. The total world production of fatty alcohols is over 2 million tons with approximately equal parts derived from fossil oil and from plant oils or animal fats. Due to the environmental impact of these production methods, there is an interest in alternative methods for fatty alcohol production via microbial fermentation using cheap renewable feedstocks. In this study, we aimed to obtain a better understanding of how fatty alcohol biosynthesis impacts the host organism, baker’s yeast Saccharomyces cerevisiae or oleaginous yeast Yarrowia lipolytica. Producing and non-producing strains were compared in growth and nitrogen-depletion cultivation phases. The multi-omics analysis included physiological characterization, transcriptome analysis by RNAseq, ¹³Cmetabolic flux analysis, and intracellular metabolomics. Both species accumulated fatty alcohols under nitrogen-depletion conditions but not during growth. The fatty alcohol–producing Y. lipolytica strain had a higher fatty alcohol production rate than an analogous S. cerevisiae strain. Nitrogen-depletion phase was associated with lower glucose uptake rates and a decrease in the intracellular concentration of acetyl–CoA in both yeast species, as well as increased organic acid secretion rates in Y. lipolytica. Expression of the fatty alcohol–producing enzyme fatty acyl–CoA reductase alleviated the growth defect caused by deletion of hexadecenal dehydrogenase encoding genes (HFD1 and HFD4) in Y. lipolytica. RNAseq analysis showed that fatty alcohol production triggered a cell wall stress response in S. cerevisiae. RNAseq analysis also showed that both nitrogen-depletion and fatty alcohol production have substantial effects on the expression of transporter encoding genes in Y. lipolytica. In conclusion, through this multi-omics study, we uncovered some effects of fatty alcohol production on the host metabolism. This knowledge can be used as guidance for further strain improvement towards the production of fatty alcohols.

Microbial carbon dioxide assimilation and conversion to chemical platform molecules has the potential to be developed as economic, sustainable processes. The carbon dioxide assimilation can proceed by a variety of natural pathways and recently even synthetic CO2 fixation routes have been designed. Early assessment of the performance of the different carbon fixation alternatives within biotechnological processes is desirable to evaluate their potential. Here we applied stoichiometric metabolic modeling based on physiological and process data to evaluate different process variants for the conversion of C1 carbon compounds to the industrial relevant platform chemical succinic acid. We computationally analyzed the performance of cyanobacteria, acetogens, methylotrophs, and synthetic CO2 fixation pathways in Saccharomyces cerevisiae in terms of production rates, product yields, and the optimization potential. This analysis provided insight into the economic feasibility and allowed to estimate the future industrial applicability by estimating overall production costs. With reported, or estimated data of engineered or wild type strains, none of the simulated microbial succinate production processes showed a performance allowing competitive production. The main limiting factors were identified as gas and photon transfer and metabolic activities whereas metabolic network structure was not restricting. In simulations with optimized parameters most process alternatives reached economically interesting values, hence, represent promising alternatives to sugar-based fermentations.

Background: Ogataea polymorpha is a thermotolerant, methylotrophic yeast with significant industrial applications. While previously mainly used for protein synthesis, it also holds promise for producing platform chemicals. O. polymorpha has the distinct advantage of using methanol as a substrate, which could be potentially derived from carbon capture and utilization streams. Full development of the organism into a production strain and estimation of the metabolic capabilities require additional strain design, guided by metabolic modeling with a genome-scale metabolic model. However, to date, no genome-scale metabolic model is available for O. polymorpha. Results: To overcome this limitation, we used a published reconstruction of the closely related yeast Komagataella phaffii as a reference and corrected reactions based on KEGG and MGOB annotation. Additionally, we conducted phenotype microarray experiments to test the suitability of 190 substrates as carbon sources. Over three-quarter of the substrate use was correctly reproduced by the model and 27 new substrates were added, that were not present in the K. phaffii reference model. Conclusion: The developed genome-scale metabolic model of O. polymorpha will support the engineering of synthetic metabolic capabilities and enable the optimization of production processes, thereby supporting a sustainable future methanol economy.

The stressosome is a bacterial signalling complex that responds to environmental changes by initiating a protein partner switching cascade, which leads to the release of the alternative sigma factor, σ^B. Stress perception increases the phosphorylation of the stressosome sensor protein, RsbR, and the scaffold protein, RsbS, by the protein kinase, RsbT. Subsequent dissociation of RsbT from the stressosome activates the σ^B cascade. However, the sequence of physical events that occur in the stressosome during signal transduction is insufficiently understood. Here, we use computational modelling to correlate the structure of the stressosome with the efficiency of the phosphorylation reactions that occur upon activation by stress. In our model, the phosphorylation of any stressosome protein is dependent upon its nearest neighbours and their phosphorylation status. We compare different hypotheses about stressosome activation and find that only the model representing the allosteric activation of the kinase RsbT, by phosphorylated RsbR, qualitatively reproduces the experimental data. Our simulations and the associated analysis of published data support the following hypotheses: (i) a simple Boolean model is capable of reproducing stressosome dynamics, (ii) different stressors induce identical stressosome activation patterns, and we also confirm that (i) phosphorylated RsbR activates RsbT, and (ii) the main purpose of RsbX is to dephosphorylate RsbS-P.

In this article we present and test a strategy to integrate, in a sequential manner, sensitivity analysis, bifurcation analysis and predictive simulations. Our strategy uses some of these methods in a coordinated way such that information, generated in one step, feeds into the definition of further analyses and helps refining the structure of the mathematical model. The aim of the method is to help in the designing of more informative predictive simulations, which focus on critical model parameters and the biological effects of their modulation. We tested our methodology with a multilevel model, accounting for the effect of erythropoietin (Epo)-mediated JAK2-STAT5 signalling in erythropoiesis. Our analysis revealed that time-delays associated with the proliferation–differentiation process are critical to induce pathological sustained oscillations, whereas the modulation of time-delays related to intracellular signalling and hypoxia-controlled physiological dynamics is not enough to induce self-oscillations in the system. Furthermore, our results suggest that the system is able to compensate (through the physiological-level feedback loop on hypoxia) the partial impairment of intracellular signalling processes (downregulation or overexpression of Epo receptor complex and STAT5), but cannot control impairment in some critical physiological-level processes, which provoke the emergence of pathological oscillations.