An outline of the project is provided in Fig. 1 and it will be comprised of four parts:
1) reconstruction of bone and muscle models;
2) quantification of diets;
3) development of a simulation toolbox;
4) use of GEMs and quantitative diets for simulation of the correlation between diet and human metabolism.
I)Reconstruction of simulation-ready GEMs for bone and muscle tissue
To date, we have reconstructed simulation-ready GEMs for human adipose and liver tissues and in connection with this project, we will reconstruct simulation-ready models for bone and muscle tissue. These models will be connected to each other using circulation models thus paving the way for studying whole body metabolism.
II) Development of diet algorithm and measuring the constituents of the diet
Diet has a major impact on the human physiology and gut microbiota. Therefore, measuring the content of diet (e.g. carbohydrates, amino acids and lipids) is essential for the assessment of the potential therapeutic and preventive effects of food intake. First, we will develop an algorithm for quantifying the constituents of 100 selected foods. The contents of the diet will be used as an input for our models and the effect of the American, European and Malawian diets and RUTF will be simulated for children and mothers.
III) Development of a toolbox for simulations
We will also develop novel methods for the simulation of the metabolic activity in different tissues. This will involve identification of different metabolic operations Such simulations will provide essential input to model the interactions between metabolically active tissues. In this context we will develop algorithms for the application of GEMs and create simulation tools that will enable integrative analysis of clinical data. In order to study the interactions between the tissues, we will generate a toolbox that can analyze the GEMs by assigning the right interactions between tissues and find the optimum solutions for carbohydrates, short chain fatty acids and amino acids profiles for different diets. We will do this based on single or multi-objective function e.g. maximization of biomass or/and minimization of total energy dissipation of ecosystems. Growth will be used as an objective function during the development of the child with and without malnutrition. The dissipation of energy will be applied as a global objective function to model the energy metabolism of mother before and after pregnancy since the growth cannot be used as an objective function for adult human tissues. The minimization of the energy dissipation for host models and defining a global objective function based on the local ones will allow for coordinating the interactions and eventually predicting the overall flux distribution in mother's body. This approach may assist to generate new hypothesis about the contribution of single tissue in the progression of malnutrition.
IV) Data integration, clinical studies and identification of improved diets
For this study we will use existing data from clinical studies and to be provided by BMGF. In case of limited data we will use data from studies of the cachexia syndrome, and we will initially collect data from these studies and evaluate their use for initial model simulations of the how different tissues alter their metabolism in response to severe weight loss. In the long run we will need high-quality clinical data from nutritional studies. These data should at least consist of high-throughput metabolome analysis on blood, feces, and urine samples for detecting significant changes in the metabolite levels and metagenome sequencing of fecal samples. Preferentially we would also obtain RNAseq data from biopsies taken from skeletal muscle and/or subcutaneous adipose tissue in children having malnutrition and responding differently to diets used to treat their malnutrition. Data on metabolite concentrations from blood, feces and urine samples collected from child and mother with and without malnutrition using untargeted metabolomics analysis will enable us to improve the functional accuracy of the reconstructed models. Furthermore, RNAseq data will allow us to refine the models for simulation of muscle and adipose energy metabolism. We will preferentially perform clinical studies using American, European and Malawian diets and RUTF. During this pilot-project we propose to develop a detailed plan for a clinical study that are closely linked with the modeling objectives as one of the first parts of the project, and this will be done in close collaboration with BMGF.
FIGURE 1 | An overview of human genome-scale metabolic models (GEMs) and their proposed applications.
(A) A metabolic network is in simple terms a list of the chemical reactions taking place in a cell. These reactions can be grouped into pathways and associated with a particular cellular compartment (e.g., mitochondria). Metabolites can be passed between compartments through transport reactions. Each reaction can be associated to its corresponding enzyme-coding genes, and together all the reactions provide a network structure connecting metabolites, reactions and genes.
(B) The metabolic network can be represented mathematically by the stoichiometric matrix, S, containing the stoichiometric coefficients of the metabolites (rows) taking part in each reaction (columns). Under the constraint based modeling framework it is assumed that the metabolite concentrations are unchanged (Sv = 0). Further on, additional constraints can be put on the flux vector, v, to find capable and probable flux distributions. Alternatively, flux balance analysis (FBA) can be used to find a flux vector that optimizes an objective function (e.g., maximize ATP production).
(C) GEMs have been used to study obesity- and diabetes-related conditions, and will here be used for studying malnutrition. Clinical data can be used to construct context specific GEMs from generic ones. This type of data can also be integrated and analyzed, in combination or separately, with the GEMs, in a topological or simulation based manner.
(D) This enables e.g., the identification of transcriptionally affected reactions and pathways as well as metabolic hotspots, or the comparison of simulation results in terms of network capabilities.