Optimization methods and objective functions for analysis of metabolic network models

Computational studies of metabolism aim to systematically analyze the metabolic behaviour of biological systems in different conditions. Genome-scale metabolic network models (GEMs) capture the connection between elements of the network by applying stoichiometric balances while taking into account gene-protein-reaction associations. Several methods have already been developed for the analysis of these models. These methods are mainly used to optimize a single or combination of objective functions subject to different constraints. In this thesis, we have summarized the optimization methods and objective functions by classifying them based on biological and mathematical features. Particularly, we suggest reformulations to convert some of the complex optimization classes to simpler ones. One of the reformulations is the conversion of mixed-integer linear fractional programming (MILFP) to mixed-integer linear programming (MILP). We show that this conversion is useful in studying coupling relationships in thermodynamically constrained metabolic network models. Coupling determines how different components of the network such as metabolites or reactions are interrelated. Particularly, flux Coupling Analysis (FCA) is a method for evaluating the dependencies between metabolic reactions. In FCA, two reactions are considered as coupled if the activity of one, constrains the activity of the other. So far, FCA has been used for analyzing metabolic reactions in flux-balanced models. In this work, we developed a new formulation, Thermodynamic Flux Coupling Analysis (TFCA), which calculates flux couplings of metabolic models that are subjected to thermodynamic constraints. With TFCA, we show that adding thermodynamic constraints can significantly change the coupling relationship of reactions of the network. Moreover, we show that calculating coupling relations helps in reducing the number of combinations of bidirectional reactions (BDRs), which in turn will facilitate the analysis of the metabolic network. In addition to proposing several mathematical reformulations to gain global optimality, we also addressed the issue of finding the proper cellular objective function in different conditions of cellular metabolism. This is not always straight-forward, since the metabolic activities of some organisms are not well-characterized, e.g., metabolism of dormancy phase in some bacteria and parasites. In this thesis, we studied the metabolic behaviour of dormant malaria parasite using genome-scale model of Plasmodium falciparum. We examined known and novel objective functions and scored them based on the modelâs consistency with experimental gene expression data. Our results suggested that minimizing energy dissipation can best describe the metabolic activities of the malaria parasites in the dormancy phase. In the last chapter, we focus on studying another poorly characterized metabolic system that is the process of iron reduction in Clostridium acetobutylicum. Research has shown that this organism can reduce Fe(III), but the mechanism behind this reduction is yet to be identified. In this thesis, we analyzed the metabolism of C. acetobutylicum using its reconstructed genome-scale metabolic network model and experimental transcriptomics data in the presence or absence of Fe(III). By performing several computational studies, we suggested that NAD(P) is involved in the reduction of iron and is the potential physiological electron donor to Fe(III).

Discovery and Evaluation of Novel Pathways for Production of the Second Generation of Biofuels

Meriç Ataman, Noushin Hadadi, Vassily Hatzimanikatis, Ljubisa Miskovic, Milenko Tokic

In an effort of overcoming the limited availability of fossil energy resources the focus of the research and development in the area of biofuels has moved towards developing the second generation of fuels that should be produced via microbial fermentation. The idea is to use as a feedstock inexpensive and abundant waste materials such as lignocellulosic biomass. The second generation biofuels should satisfy several criteria such as lower emission, higher energy density and should be less corrosive to engines. However, currently used industrial workhorses such as E. coli and S. cerevisiae do not produce many of these molecules naturally. Furthermore, for majority of these molecules there are no known biochemical pathways. The need for discovery of novel biosynthetic pathways towards desired molecules sparked the development of computational tools that are capable of reconstructing feasible reaction steps between a given set of starting compounds (precursors) and a molecule of interest (a target compound). In this study, we performed the retrobiosynthesis analysis for all the candidates applying Biochemical Network Integrated Computational Explorer (BNICE.ch) and we reconstructed the metabolic network for 69 fuel compounds. We analyzed compounds (biological or chemical) and reactions one-step away from these target molecules and the appearance of the potential substrates in their reconstructed metabolic network. Based on the results of this analysis, coupled with other criteria such as the Gibes free energy of formation, combustion potential and expert opinion, we chose the 5 highest ranked fuel candidates for further pathway reconstruction studies. For these 5 fuel candidates we have created 1.8 million de novo pathways with different number of reaction steps for further evaluation. We have then embedded the reconstructed pathways in the genome-°©‐scale models of the two chassis organisms E. coli and P. putida and we performed the pathway evaluation with respect to their maximum theoretical yield using two methods (i) Flux Balance Analysis (FBA), and (ii) Thermodynamic-based Flux Analysis (TFA). All the pathways that did not satisfy the thermodynamic constraints were rejected. Our results indicated that in the evaluation and ranking of the pathways the thermodynamics is a necessary consideration. Specifically, we showed that the majority of the pathways that were determined as feasible based on FBA, they were not feasible based on TFA. Moreover, we also showed that some pathways that were determined as feasible in both organisms based on FBA, they were not feasible in both of them based on TFA. Also, the yield of the feasible pathways can differ depending on the organism. These results demonstrate how TFA can also guide the selection of a suitable chassis organism. For all discovered reactions in the feasible pathways we have also identified known enzymes from the KEGG database with the closest reaction similarity that could be engineered for the implementation of novel reactions. This study shows the full potential of computational tools for discovery and design of novel synthetic pathways and their relevance for future developments in the area of metabolic engineering and synthetic biology.

2016

Bioenergetics-based modeling of Plasmodium falciparum metabolism reveals its essential genes, nutritional requirements, and thermodynamic bottlenecks

Meriç Ataman, Vassily Hatzimanikatis, Dominique Soldati-Favre, Stepan Tymoshenko

Novel antimalarial therapies are urgently needed for the fight against drug-resistant para- sites. The metabolism of malaria parasites in infected cells is an attractive source of drug targets but is rather complex. Computational methods can handle this complexity and allow integrative analyses of cell metabolism. In this study, we present a genome-scale metabolic model (iPfa) of the deadliest malaria parasite, Plasmodium falciparum, and its thermody- namics-based flux analysis (TFA). Using previous absolute concentration data of the intraer- ythrocytic parasite, we applied TFA to iPfa and predicted up to 63 essential genes and 26 essential pairs of genes. Of the 63 genes, 35 have been experimentally validated and reported in the literature, and 28 have not been experimentally tested and include previously hypothesized or novel predictions of essential metabolic capabilities. Without metabolomics data, four of the genes would have been incorrectly predicted to be non-essential. TFA also indicated that substrate channeling should exist in two metabolic pathways to ensure the thermodynamic feasibility of the flux. Finally, analysis of the metabolic capabilities of P. fal- ciparum led to the identification of both the minimal nutritional requirements and the genes that can become indispensable upon substrate inaccessibility. This model provides novel insight into the metabolic needs and capabilities of the malaria parasite and highlights metabolites and pathways that should be measured and characterized to identify potential thermodynamic bottlenecks and substrate channeling. The hypotheses presented seek to guide experimental studies to facilitate a better understanding of the parasite metabolism and the identification of targets for more efficient intervention.

Public Library of Science2017

Analysis of cellular and drug action mechanisms in the malaria parasites using systems biology approaches

Anush Chiappino-Pepe

Malaria puts at risk of infection half of the world population and still kills around half a million people every year. The rapid and worrisome rise of drug-resistant malaria parasites hinders the treatment of malaria and calls for a global collaboration to eradicate this disease. The technological advances have allowed sequencing the genome and measuring metabolites, proteins, and mRNAs of the parasites at a high-throughput, which has provided insights into the biological processes of these organisms. However, many aspects of the biology of the malaria parasites remain hidden and hinder the development of effective and selective antimalarial therapies. The understanding of the cell function as a whole rather than the study of its components in isolation will enable the development of strategies to engineer and target the cell function. Only computational tools can provide a comprehensive understanding of the cellular function in different contexts and upon perturbation. Moreover, computational methods provide testable hypotheses that can guide and accelerate experimental discoveries. Metabolism is to date the most characterized biological process in the cell and is an attractive source of drug targets. Metabolic networks encompass all available information on the biochemistry of an organism and represent an attractive scaffold to integrate omics data and analyze complex metabolic processes. In this thesis, we present the metabolic models of two malaria parasites, i.e., Plasmodium falciparum and Plasmodium berghei, and a set of mathematical and computational methods for the comprehensive and unbiased analysis of metabolic networks. We focus on the bioenergetics-based and metabolic control analysis of the malaria parasitesâ metabolism to find drug targets and investigate drug action mechanisms. Firstly, we develop computational methods that follow principles from network thermodynamics to identify essential metabolic functions, thermodynamic bottlenecks, and nutritional requirements of P. falciparum. Secondly, we define a mathematical framework for Phenotype Mapping (PhenoMapping), which associates a cellular phenotype with the underlying cellular processes and conditions. We applied PhenoMapping to investigate the PlasmoGEM phenotypes obtained for P. berghei in the blood stages of its life cycle. This analysis enables the identification of non-conditionally and conditionally essential metabolic functions and the development of context-specific metabolic models. Thirdly, we present a Mixed Integer Synthetic Lethal Enumeration (MISyLE) method for the optimal identification of essential and all redundant (synthetic lethal) sets of reactions and genes for a cellular phenotype. Fourthly, we investigate the effect of antimalarial-based therapies with one or multiple antifolates in the metabolism of P. falciparum in the blood stages, and we examine new drug targeting strategies that might maximize the impact on the intraerythrocytic development of the parasites. Fifthly, we discuss a systems biology approach to accelerate our understanding of biological systems based on the disagreements between in silico and in vivo observations. Finally, we suggest strategies to use the models and computational frameworks developed in this thesis to further investigate host-pathogen interactions, drug action mechanisms, and dormancy in the malaria parasites and other organisms.

EPFL2018