An important challenge of contemporary biology and medicine is to establish the relationship between disease phenotypes and disruptions in the underlying cellular functions. In the past decades, substantial efforts have been devoted to a gene-based approach, identifying the specific genetic defects that, together with single nucleotide polymorphisms and copy number variations, predispose an individual to disease. However, it is increasingly recognized that such an approach, although somewhat successful, is far from sufficient. This is especially because most cellular components exert their function through intricate networks of regulatory, metabolic, and protein interactions. Consequently, the impact of different (and often disease-causing) genetic and epigenetic variations are not restricted but are spread in the intracellular network, affecting the activity and/or function of gene products that otherwise carry no defects. Because of these complex interdependencies among a cell’s molecular components, the functional and causal relationships between genotype and disease phenotype is extremely complex, and difficult to unravel.
It is now a recognized fact that most human diseases derive either from failure in metabolic enzymes, or are at least propagated by perturbations in metabolic pathways. Therefore, interrogation of metabolic pathways offers a viable strategy for exploring disease mechanisms. This is particularly true because metabolism is the only system that can be fully simulated at genome scale. Metabolic systems biology offers powerful abstraction tools to simulate all known metabolic reactions in a cell, therefore providing a snapshot that is close to its observable phenotype. Though yet far from complete, metabolic reconstructions are now sufficiently mature to provide testable computational predictions. Importantly, these reconstructions are arguably more complete than those for signal transduction, gene regulation, or any other type of cellular network. Indeed, correlations between enzymes – based on the static network structure and shared metabolites – have recently been mapped to correlations in disease phenotypes.
Exploiting metabolic flux as an approach to disease-specific drug target discovery.
Metabolism represents the ‘sharp end’ of systems biology because changes in metabolite concentrations are necessarily amplified relative to changes in the transcriptome, proteome and enzyme activities. Modeling metabolic networks has an advantage over transcriptomic or proteomic networks because, although changes in the activities of individual enzymes tend to have rather small effects on metabolic fluxes, they can and do have very large effects on metabolite concentrations (i.e. the metabolome). Thus, the metabolome serves to amplify possibly immeasurably small changes in the transcriptome and the proteome, even when derived from minor changes in the genome
Over the past decade, several groups have been independently developing models of the human metabolic network, both at a generalized level and in tissue-specific forms. The most significant development, however, is the recent description of a highly curated and semantically annotated model of the human metabolic network, termed Recon2 (http://humanmetabolism.org/). Many major groups collaborated in this endeavor to provide a carefully and manually constructed/curated network, consisting of some 1789 enzyme-encoding genes, 7440 reactions and 2626 unique metabolites distributed over eight cellular compartments. Armed with such metabolic network models, one can now predict metabolic fluxes directly. This can be done either in a ‘forward’ direction wherein, given the network, starting concentrations of enzymes and metabolites, and rate equations one can predict the fluxes. Conversely, an ‘inverse’ direction can also be employed where, given the fluxes and concentrations, one can try to predict the enzyme concentrations and kinetic parameters that would account for them. Iterative use of both kinds of knowledge are also possible. In this context, a complementary discovery of importance is that a single transcriptome experiment, serving as a surrogate for fluxes through individual steps, provides a large constraint on possible models, and directly predicts – in a numerically tractable way and with much improved accuracy – the metabolic fluxes. Work continues on related strategies that exploit modern advances in ‘omics’ and network biology to limit the search space in constraint-based metabolic modelling.
To translate progress in metabolic network modelling into benefits for drug discovery, what is now needed is a ‘systems pharmacology’ approach where multiple binding targets are chosen rationally and simultaneously. By combining this with other approaches such as phenotypic screening, and ‘omics’ based strategies, one can expect considerable improvements in the rate of discovery of safe and effective drugs
Presently, the pharmaceutical industry is being plagued by increasing research and development costs on the one hand, while approximately 90% of drug candidates tested fail during clinical development. Most of these failures are simply due to a lack of efficacy, and progressing candidates through the drug development value chain without a full understanding of their impact on the metabolic regulation of disease. Integrating targeted hypothesis-driven metabolomic approaches with preclinical and clinical drug development may facilitate a redesign of scientific paradigms within the current business model of the pharmaceutical industry.