DFBA-LQR: An Optimal Control Approach to Flux Balance Analysis

Microbial interactions from a new perspective: reinforcement learning reveals new insights into microbiome evolution

MOTIVATION

Microbes are essential part of all ecosystems, influencing material flow and shaping their surroundings. Metabolic modeling has been a useful tool and provided tremendous insights into microbial community metabolism. However, current methods based on flux balance analysis (FBA) usually fail to predict metabolic and regulatory strategies that lead to long-term survival and stability especially in heterogenous communities.

RESULTS

Here, we introduce a novel reinforcement learning algorithm, Self-Playing Microbes in Dynamic FBA, which treats microbial metabolism as a decision-making process, allowing individual microbial agents to evolve by learning and adapting metabolic strategies for enhanced long-term fitness. This algorithm predicts what microbial flux regulation policies will stabilize in the dynamic ecosystem of interest in the presence of other microbes with minimal reliance on predefined strategies. Throughout this article, we present several scenarios wherein our algorithm outperforms existing methods in reproducing outcomes, and we explore the biological significance of these predictions.

AVAILABILITY AND IMPLEMENTATION

The source code for this article is available at: https://github.com/chan-csu/SPAM-DFBA.

Synthetic biology and regulatory networks: where metabolic systems biology meets control engineering

Metabolic pathways can be engineered to maximize the synthesis of various products of interest. With the advent of computational systems biology, this endeavour is usually carried out through in silico theoretical studies with the aim to guide and complement further in vitro and in vivo experimental efforts. Clearly, what counts is the result in vivo, not only in terms of maximal productivity but also robustness against environmental perturbations. Engineering an organism towards an increased production flux, however, often compromises that robustness. In this contribution, we review and investigate how various analytical approaches used in metabolic engineering and synthetic biology are related to concepts developed by systems and control engineering. While trade-offs between production optimality and cellular robustness have already been studied diagnostically and statically, the dynamics also matter. Integration of the dynamic design aspects of control engineering with the more diagnostic aspects of metabolic, hierarchical control and regulation analysis is leading to the new, conceptual and operational framework required for the design of robust and productive dynamic pathways.

Resuscitation of ischemic donor livers with normothermic machine perfusion: a metabolic flux analysis of treatment in rats

Normothermic machine perfusion has previously been demonstrated to restore damaged warm ischemic livers to transplantable condition in animal models. However, the mechanisms of recovery are unclear, preventing rational optimization of perfusion systems and slowing clinical translation of machine perfusion. In this study, organ recovery time and major perfusate shortcomings were evaluated using a comprehensive metabolic analysis of organ function in perfusion prior to successful transplantation. Two groups, Fresh livers and livers subjected to 1 hr of warm ischemia (WI) received perfusion for a total preservation time of 6 hrs, followed by successful transplantation. 24 metabolic fluxes were directly measured and 38 stoichiometrically-related fluxes were estimated via a mass balance model of the major pathways of energy metabolism. This analysis revealed stable metabolism in Fresh livers throughout perfusion while identifying two distinct metabolic states in WI livers, separated at t = 2 hrs, coinciding with recovery of oxygen uptake rates to Fresh liver values. This finding strongly suggests successful organ resuscitation within 2 hrs of perfusion. Overall perfused livers regulated metabolism of perfusate substrates according to their metabolic needs, despite supraphysiological levels of some metabolites. This study establishes the first integrative metabolic basis for the dynamics of recovery during perfusion treatment of marginal livers. Our initial findings support enhanced oxygen delivery for both timely recovery and long-term sustenance. These results are expected to lead the optimization of the treatment protocols and perfusion media from a metabolic perspective, facilitating translation to clinical use.

Development of Metabolic Indicators of Burn Injury: Very Low Density Lipoprotein (VLDL) and Acetoacetate Are Highly Correlated to Severity of Burn Injury in Rats

Hypermetabolism is a significant sequela to severe trauma such as burns, as well as critical illnesses such as cancer. It persists in parallel to, or beyond, the original pathology for many months as an often-fatal comorbidity. Currently, diagnosis is based solely on clinical observations of increased energy expenditure, severe muscle wasting and progressive organ dysfunction. In order to identify the minimum number of necessary variables, and to develop a rat model of burn injury-induced hypermetabolism, we utilized data mining approaches to identify the metabolic variables that strongly correlate to the severity of injury. A clustering-based algorithm was introduced into a regression model of the extent of burn injury. As a result, a neural network model which employs VLDL and acetoacetate levels was demonstrated to predict the extent of burn injury with 88% accuracy in the rat model. The physiological importance of the identified variables in the context of hypermetabolism, and necessary steps in extension of this preliminary model to a clinically utilizable index of severity of burn injury are outlined.

Optimization-based metabolic control analysis

In this work, a novel optimization-based metabolic control analysis (OMCA) method is introduced for reducing data requirement for metabolic control analysis (MCA). It is postulated that using the optimal control approach, the fluxes in a metabolic network are correlated to metabolite concentrations and enzyme activities as a state-feedback control system that is optimal with respect to a homeostasis objective. It is then shown that the optimal feedback gains are directly related to the elasticity coefficients (ECs) of MCA. This approach requires determination of the relative "importance" of metabolites and fluxes for the system, which is possible with significantly reduced experimental data, as compared with typical MCA requirements. The OMCA approach is applied to a top-down control model of glycolysis in hepatocytes. It is statistically demonstrated that the OMCA model is capable of predicting the ECs observed experimentally with few exceptions. Further, an OMCA-based model reconciliation study shows that the modification of four assumed stoichiometric coefficients in the model can explain most of the discrepancies, with the exception of elasticities with respect to the NADH/NAD ratio.

Expanding a dynamic flux balance model of yeast fermentation to genome-scale

Background

Yeast is considered to be a workhorse of the biotechnology industry for the production of many value-added chemicals, alcoholic beverages and biofuels. Optimization of the fermentation is a challenging task that greatly benefits from dynamic models able to accurately describe and predict the fermentation profile and resulting products under different genetic and environmental conditions. In this article, we developed and validated a genome-scale dynamic flux balance model, using experimentally determined kinetic constraints.

Results

Appropriate equations for maintenance, biomass composition, anaerobic metabolism and nutrient uptake are key to improve model performance, especially for predicting glycerol and ethanol synthesis. Prediction profiles of synthesis and consumption of the main metabolites involved in alcoholic fermentation closely agreed with experimental data obtained from numerous lab and industrial fermentations under different environmental conditions. Finally, fermentation simulations of genetically engineered yeasts closely reproduced previously reported experimental results regarding final concentrations of the main fermentation products such as ethanol and glycerol.

Conclusion

A useful tool to describe, understand and predict metabolite production in batch yeast cultures was developed. The resulting model, if used wisely, could help to search for new metabolic engineering strategies to manage ethanol content in batch fermentations.