Abstract
Metabolic modeling enables the study of human metabolism in healthy and in diseased conditions, e.g., the prediction of new drug targets and biomarkers for metabolic diseases. To accurately describe blood and urine metabolite dynamics, the integration of multiple metabolically active tissues is necessary. We developed a dynamic multi-tissue model, which recapitulates key properties of human metabolism at the molecular and physiological level based on the integration of transcriptomics data. It enables the simulation of the dynamics of intra-cellular and extra-cellular metabolites at the genome scale. The predictive capacity of the model is shown through the accurate simulation of different healthy conditions (i.e., during fasting, while consuming meals or during exercise), and the prediction of biomarkers for a set of Inborn Errors of Metabolism with a precision of 83%. This novel approach is useful to prioritize new biomarkers for many metabolic diseases, as well as for the integration of various types of personal omics data, towards the personalized analysis of blood and urine metabolites.
Highlights
A study by Hyötyläinen et al.[11] suggested to relax these hard constraints by using a quadratic programming approach in which hard bounds are replaced by penalties for data violation. This approach enables the direct use of data in a model, while contradicting bounds are relaxed in an automatic fashion, with many smaller violations preferred to large individual violations
We further applied stronger perturbations such as the knock-out of genes-associated with Inborn Errors of Metabolism (IEMs) to investigate their impact on blood metabolites levels, and on urine excretion rates. 90% of the investigated metabolic changes occurring during incremental exercise could be matched with previous data, and the prediction of blood amino acids biomarkers had a precision of 83%
Three constraint-based metabolic tissue models (Fig. 1), liver, muscle and adipose tissue were reconstructed from Recon2.043 and tissue specific transcriptomic data, and integrated to create a multi-tissue model that contains a total of 7251 reactions, and 5311 metabolites
Summary
Genome scale reconstructions of the human metabolism have become ever more sophisticated in the recent years (EHMN1, Recon[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32–5], and HMR1-26,7). A study by Hyötyläinen et al.[11] suggested to relax these hard constraints by using a quadratic programming approach in which hard bounds are replaced by penalties for data violation This approach enables the direct use of data in a model, while contradicting bounds are relaxed in an automatic fashion, with many smaller violations preferred to large individual violations. This approach still does not address another issue with multi-tissue models of higher organisms. Temporary perturbations of metabolite levels occur primarily due to nutrient uptake or food scarcity. We further applied stronger perturbations such as the knock-out of genes-associated with IEMs to investigate their impact on blood metabolites levels, and on urine excretion rates. 90% of the investigated metabolic changes occurring during incremental exercise could be matched with previous data, and the prediction of blood amino acids biomarkers had a precision of 83%
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
