Systematic reduction of genome-scale models for the study of metabolic phenotypes of human cells

In the last years, the analysis of cellular metabolism has sparked new interest in systems biology and metabolic modelling. In particular, modelling the phenotypes of healthy and diseased cells will help to understand the main metabolic characteristics of disease development and progression. It will also be key to design more effective therapies. The reconstruction of genome-scale models (GEMs) enables the computation of phenotypic traits based on the genetic composition of a target organism. To overcome the well-known challenges when working with large networks, we generate systematically reduced models around specific subsystems. Within this framework, we curate the GEMs to include the thermodynamic properties of the network metabolites and reactions. Furthermore, we consider the composition and utilization of the extracellular-medium metabolites and the synthesis of the biomass precursor metabolites. The reduced models can be used for a broad range of applications extending from omics data integration to kinetic models. We demonstrate here that the study of reduced human GEMs can provide insight and a systematic framework to compare different versions of GEMs (Recon 2 versus Recon 3) as well as different cellular physiologies, such as diseased versus healthy phenotypes.

Studying cancer cells phenotypes integrating omics data in human reduced genome-scale metabolic models.

Vassily Hatzimanikatis, Maria Masid Barcon, Vikash Kumar Pandey

Cancer is a leading cause of death in the world, and the mechanisms underlying this disease are still not completely understood. In the last decades, altered tumour metabolism has been recognized as a hallmark of cancer. This has created a resurgence of interest in the field of systems biology and metabolic modelling to analyse and understand the metabolic changes occurring in cancer cells. Modelling the different phenotypes of healthy and cancer cells will help to make predictions to create effective therapies to prevent, diagnose and treat cancer. The reconstruction of genome-scale models (GEMs) enables the computation of phenotypic traits based on the genetic composition of a target organism. Therefore, we use the human genome-scale model Recon 2 to build cancer cell-specific human GEMs by integrating experimental data (fluxomics, metabolomics, genomics, transcriptomics) and thermodynamic data. To overcome the well-known challenges when working with large networks, we generate systematically reduced models around specific subsystems, considering the composition and usage of the extracellular medium metabolites, and the biosynthesis of the biomass precursor metabolites. Furthermore, we apply pathway enrichment to study the regulation of the pathways under study. The reduced models can be used for a broad range of applications ranging from omics data integration to kinetic models. The proposed pipeline will enhance the comparison and understanding, at the stoichiometric and kinetic level, of the main metabolic differences that emerge in cancer development and progression. Furthermore, predicting the network responses will help to design experiments to find new targets for therapies and drugs.

2018

Mapping and Analysis of the Complexity of Lipid Metabolism for Applications in Health and Sustainability

Vassily Hatzimanikatis, Sofia Tsouka

Metabolism is a network of biotransformations that sustain life in cells. Over the past decades, metabolism has been studied in multiple contexts, ranging from medicine to industrial manufacturing. In particular, lipid metabolism has been linked to various physiopathologies, and its study could lead to new diagnostics and treatments. Metabolic networks are extremely complex and a systems approach is essential in deciphering the different levels of cellular organization and regulation that an organism possesses. Genome-scale metabolic models (GEMs) encompass all the available information for an organism, though their large size and complexity introduce a lot of uncertainty in the accuracy of predictions. GEM reductions are usually done in an ad hoc manner to produce context-specific models, and cannot serve multiple studies. The study of regulatory mechanisms requires the development of kinetic models that can accurately capture dynamic responses to perturbations. To this end, constraint-based models can be enhanced through the integration of kinetic information. Constructing consistent large-scale kinetic models is a challenging endeavor, since detailed knowledge about regulatory mechanisms and kinetic parameter values is scarce.Examining the mechanisms of enzymatic and metabolic restructuring in states of mutation could increase our fundamental understanding of how diseases evolve, and facilitate phenotype mapping. Metabolic Control Analysis (MCA) has been a well-established tool for the prediction and evaluation of genetic modification strategies. However, it fails to account for the physiological limitations of an organism and can often lead to unrealistic predictions.In this thesis, we developed computational models, tools, and methodologies to facilitate the study of lipids and their regulatory mechanisms. Firstly, we constructed and curated a metabolic model that focuses on lipid metabolism, through the integration of detailed lipid pathways into a GEM of S. cerevisiae. This model was then systematically reduced around these pathways to provide a more manageable model size for complex studies. We show that this model is as consistent and inclusive as other yeast GEMs, and can be used as a scaffold for integrating lipidomics data to improve predictions in studies of lipid-related biological functions. Secondly, we used this model as a basis to build a large-scale kinetic model. Enzymes in the lipid metabolism are typically promiscuous and multifunctional, giving rise to enzymatic coupling. To accurately capture these properties, we assigned suitable kinetic rate expressions to the reactions in the network. We generated populations of kinetic parameter sets through a sampling-based workflow and we demonstrate how the consideration of enzymatic coupling is essential for factual predictions. Thirdly, we developed a constraint-based formulation that utilizes MCA-based control coefficients for the consistent derivation of metabolic engineering strategies. We show how the parametrization and introduction of biological constraints enhances this formulation in comparison to the classical MCA approach, and we highlight its ability to generate alternative optimal strategies. Fourthly, we defined and introduced additional mathematical objectives and constraints to this formulation to enable the mapping of enzymes that are responsible for different phenotypes. Concluding, we discuss the contribution and potential applications of this thesis.

EPFL2020

Mapping and Analysis of the Complexity of Lipid Metabolism for Applications in Health and Sustainability

Vassily Hatzimanikatis, Sofia Tsouka

EPFL2020