Metabolomics and systems pharmacology: why and how to model the human metabolic network for drug discovery

A data preprocessing strategy for metabolomics to reduce the mask effect in data analysis

Highlights

Developed a data preprocessing strategy to cope with missing values and mask effects in data analysis from high variation of abundant metabolites.

A new method- ‘x-VAST’ was developed to amend the measurement deviation enlargement.

Applying the above strategy, several low abundant masked differential metabolites were rescued.

Metabolomics is a booming research field. Its success highly relies on the discovery of differential metabolites by comparing different data sets (for example, patients vs. controls). One of the challenges is that differences of the low abundant metabolites between groups are often masked by the high variation of abundant metabolites. In order to solve this challenge, a novel data preprocessing strategy consisting of three steps was proposed in this study. In step 1, a ‘modified 80%’ rule was used to reduce effect of missing values; in step 2, unit-variance and Pareto scaling methods were used to reduce the mask effect from the abundant metabolites. In step 3, in order to fix the adverse effect of scaling, stability information of the variables deduced from intensity information and the class information, was used to assign suitable weights to the variables. When applying to an LC/MS based metabolomics dataset from chronic hepatitis B patients study and two simulated datasets, the mask effect was found to be partially eliminated and several new low abundant differential metabolites were rescued.

Transparency in metabolic network reconstruction enables scalable biological discovery

Reconstructing metabolic pathways has long been a focus of active research. Now, draft models can be generated from genomic annotation and used to simulate metabolic fluxes of mass and energy at the whole-cell scale. This approach has led to an explosion in the number of functional metabolic network models. However, more models have not led to expanded coverage of metabolic reactions known to occur in the biosphere. Thus, there exists opportunity to reconsider the process of reconstruction and model derivation to better support the less-scalable investigative processes of biocuration and experimentation. Realizing this opportunity to improve our knowledge of metabolism requires developing new tools that make reconstructions more useful by highlighting metabolic network knowledge limitations to guide future research.

Current status and future prospects of mass spectrometry imaging of small molecules

In the field of small-molecule studies, vast efforts have been put forth in order to comprehensively characterize and quantify metabolites formed from complex mechanistic pathways within biochemical and biological organisms. Many technologies and methodologies have been developed to aid understanding of the inherent complexities within biological metabolomes. Specifically, mass spectroscopy imaging (MSI) has emerged as a foundational technique in gaining insight into the molecular entities within cells, tissues, and whole-body samples. In this chapter we provide a brief overview of major technical components involved in MSI, including topics such as sample preparation, analyte ionization, ion detection, and data analysis. Emerging applications are briefly summarized as well, but details will be presented in the following chapters.

Molecular phenotyping of a UK population: defining the human serum metabolome

Phenotyping of 1,200 'healthy' adults from the UK has been performed through the investigation of diverse classes of hydrophilic and lipophilic metabolites present in serum by applying a series of chromatography-mass spectrometry platforms. These data were made robust to instrumental drift by numerical correction; this was prerequisite to allow detection of subtle metabolic differences. The variation in observed metabolite relative concentrations between the 1,200 subjects ranged from less than 5 % to more than 200 %. Variations in metabolites could be related to differences in gender, age, BMI, blood pressure, and smoking. Investigations suggest that a sample size of 600 subjects is both necessary and sufficient for robust analysis of these data. Overall, this is a large scale and non-targeted chromatographic MS-based metabolomics study, using samples from over 1,000 individuals, to provide a comprehensive measurement of their serum metabolomes. This work provides an important baseline or reference dataset for understanding the 'normal' relative concentrations and variation in the human serum metabolome. These may be related to our increasing knowledge of the human metabolic network map. Information on the Husermet study is available at http://www.husermet.org/. Importantly, all of the data are made freely available at MetaboLights (http://www.ebi.ac.uk/metabolights/).

Serum ferritin is an important inflammatory disease marker, as it is mainly a leakage product from damaged cells

"Serum ferritin" presents a paradox, as the iron storage protein ferritin is not synthesised in serum yet is to be found there. Serum ferritin is also a well known inflammatory marker, but it is unclear whether serum ferritin reflects or causes inflammation, or whether it is involved in an inflammatory cycle. We argue here that serum ferritin arises from damaged cells, and is thus a marker of cellular damage. The protein in serum ferritin is considered benign, but it has lost (i.e. dumped) most of its normal complement of iron which when unliganded is highly toxic. The facts that serum ferritin levels can correlate with both disease and with body iron stores are thus expected on simple chemical kinetic grounds. Serum ferritin levels also correlate with other phenotypic readouts such as erythrocyte morphology. Overall, this systems approach serves to explain a number of apparent paradoxes of serum ferritin, including (i) why it correlates with biomarkers of cell damage, (ii) why it correlates with biomarkers of hydroxyl radical formation (and oxidative stress) and (iii) therefore why it correlates with the presence and/or severity of numerous diseases. This leads to suggestions for how one might exploit the corollaries of the recognition that serum ferritin levels mainly represent a consequence of cell stress and damage.

What would be the observable consequences if phospholipid bilayer diffusion of drugs into cells is negligible?

For drug transport across (i.e., through) an intact biological membrane, two main routes are possible: drugs may cross (i) through the phospholipid bilayer portion of the membrane, and/or (ii) via proteinaceous pores or transporters. Perhaps surprisingly, there is in fact no direct scientific evidence that the first of these takes place at any significant rate because, in the experiments performed to date, it has neither been varied as an independent variable nor measured directly as a dependent variable. Using a standard hypothetico-deductive framework, I assess the intellectual and observable consequences of assuming that, for drugs, phospholipid bilayer diffusion is negligible - 'PBIN' - (i.e., may be neglected, relative to transporter-mediated transmembrane fluxes). Predictions and postdictions of the PBIN hypothesis are not refuted by available experimental evidence.

How drugs get into cells: tested and testable predictions to help discriminate between transporter-mediated uptake and lipoidal bilayer diffusion

One approach to experimental science involves creating hypotheses, then testing them by varying one or more independent variables, and assessing the effects of this variation on the processes of interest. We use this strategy to compare the intellectual status and available evidence for two models or views of mechanisms of transmembrane drug transport into intact biological cells. One (BDII) asserts that lipoidal phospholipid Bilayer Diffusion Is Important, while a second (PBIN) proposes that in normal intact cells Phospholipid Bilayer diffusion Is Negligible (i.e., may be neglected quantitatively), because evolution selected against it, and with transmembrane drug transport being effected by genetically encoded proteinaceous carriers or pores, whose “natural” biological roles, and substrates are based in intermediary metabolism. Despite a recent review elsewhere, we can find no evidence able to support BDII as we can find no experiments in intact cells in which phospholipid bilayer diffusion was either varied independently or measured directly (although there are many papers where it was inferred by seeing a covariation of other dependent variables). By contrast, we find an abundance of evidence showing cases in which changes in the activities of named and genetically identified transporters led to measurable changes in the rate or extent of drug uptake. PBIN also has considerable predictive power, and accounts readily for the large differences in drug uptake between tissues, cells and species, in accounting for the metabolite-likeness of marketed drugs, in pharmacogenomics, and in providing a straightforward explanation for the late-stage appearance of toxicity and of lack of efficacy during drug discovery programmes despite macroscopically adequate pharmacokinetics. Consequently, the view that Phospholipid Bilayer diffusion Is Negligible (PBIN) provides a starting hypothesis for assessing cellular drug uptake that is much better supported by the available evidence, and is both more productive and more predictive.

NMR spectroscopy-based metabolic profiling of drug-induced changes in vitro can discriminate between pharmacological classes

Drug-induced changes in mammalian cell line models have already been extensively profiled at the systemic mRNA level and subsequently used to suggest mechanisms of action for new substances, as well as to support drug repurposing, i.e., identifying new potential indications for drugs already licensed for other pharmacotherapy settings. The seminal work in this field, which includes a large database and computational algorithms for pattern matching, is known as the "Connectivity Map" (CMap). However, the potential of similar exercises at the metabolite level is still largely unexplored. Only recently, the first high-throughput metabolomic assay pilot study was published, which involved screening the metabolic response to a set of 56 kinase inhibitors in a 96-well format. Here, we report results from a separately developed metabolic profiling assay, which leverages (1)H NMR spectroscopy to the quantification of metabolic changes in the HCT116 colorectal cancer cell line, in response to each of 26 compounds. These agents are distributed across 12 different pharmacological classes covering a broad spectrum of bioactivity. Differential metabolic profiles, inferred from multivariate spectral analysis of 18 spectral bins, allowed clustering of the most-tested drugs, according to their respective pharmacological class. A more-advanced supervised analysis, involving one multivariate scattering matrix per pharmacological class and using only 3 spectral bins (3 metabolites), showed even more distinct pharmacology-related cluster formations. In conclusion, this type of relatively fast and inexpensive profiling seems to provide a promising alternative to that afforded by mRNA expression analysis, which is relatively slow and costly. As also indicated by the present pilot study, the resulting metabolic profiles do not seem to provide as information-rich signatures as those obtained using systemic mRNA profiling, but the methodology holds strong promise for significant refinement.

A 'rule of 0.5' for the metabolite-likeness of approved pharmaceutical drugs

We exploit the recent availability of a community reconstruction of the human metabolic network ('Recon2') to study how close in structural terms are marketed drugs to the nearest known metabolite(s) that Recon2 contains. While other encodings using different kinds of chemical fingerprints give greater differences, we find using the 166 Public MDL Molecular Access (MACCS) keys that 90 % of marketed drugs have a Tanimoto similarity of more than 0.5 to the (structurally) 'nearest' human metabolite. This suggests a 'rule of 0.5' mnemonic for assessing the metabolite-like properties that characterise successful, marketed drugs. Multiobjective clustering leads to a similar conclusion, while artificial (synthetic) structures are seen to be less human-metabolite-like. This 'rule of 0.5' may have considerable predictive value in chemical biology and drug discovery, and may represent a powerful filter for decision making processes.

An analysis of a 'community-driven' reconstruction of the human metabolic network

Following a strategy similar to that used in baker's yeast (Herrgård et al. Nat Biotechnol 26:1155-1160, 2008). A consensus yeast metabolic network obtained from a community approach to systems biology (Herrgård et al. 2008; Dobson et al. BMC Syst Biol 4:145, 2010). Further developments towards a genome-scale metabolic model of yeast (Dobson et al. 2010; Heavner et al. BMC Syst Biol 6:55, 2012). Yeast 5-an expanded reconstruction of the Saccharomyces cerevisiae metabolic network (Heavner et al. 2012) and in Salmonella typhimurium (Thiele et al. BMC Syst Biol 5:8, 2011). A community effort towards a knowledge-base and mathematical model of the human pathogen Salmonellatyphimurium LT2 (Thiele et al. 2011), a recent paper (Thiele et al. Nat Biotechnol 31:419-425, 2013). A community-driven global reconstruction of human metabolism (Thiele et al. 2013) described a much improved 'community consensus' reconstruction of the human metabolic network, called Recon 2, and the authors (that include the present ones) have made it freely available via a database at http://humanmetabolism.org/ and in SBML format at Biomodels (http://identifiers.org/biomodels.db/MODEL1109130000). This short analysis summarises the main findings, and suggests some approaches that will be able to exploit the availability of this model to advantage.