More on this book
Kindle Notes & Highlights
Read between
December 4, 2019 - May 22, 2020
Regardless of the actual number, the goal of “metabolomics” is to enable robust measurement of endogenous small molecules for the determination of metabolic phenotypes.
In particular, differences in polarity among groups or families of metabolites demand that different extraction procedures be used during the preparation of analytical samples.
The analytical tools most commonly used in metabolomics, nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS), have linear dynamic ranges for analyses that span 3 to 5 orders of magnitude, but the concentration range of metabolites in any sample will exceed this range.
A third challenge arises from differences in chemical stability and, consequently, differences in rates of decomposition among metabolites, both in intact biological samples and after metabolite extraction procedures.
Deeper coverage is often limited by the amount of biological sample that is available and practical considerations of time and cost.
The analytical tools most commonly used in metabolomics are NMR spectroscopy, liquid and gas chromatography (LC, GC), and MS.
Targeted metabolomics is typically limited to a set of known, chemically related metabolites.
In contrast, untargeted metabolomics measures all endogenous metabolic signals in a biological sample. These methods are necessarily broader in order to capture the larger range of metabolites, but they result in reduced sensitivity.
Although this process is tedious, an untargeted approach allows for the identification of unknown metabolites that may, in turn, have a significant impact on disease processes. Therefore, the potential for novel scientific findings is higher, although further scrutiny of the results is necessary.
Data cleaning and processing can be completed using several resources, one of the most popular being MetaboAnalyst (Xia, Mandal, et al. 2012).
Quality-control checking allows for the examination of systematic errors, batch effects, sample decay, and outlying observations. Other analysis programs are also available through R, including the Metabolite Automatic Identification Toolkit (MAIT), which can identify and annotate MS peaks as well as perform initial statistical analyses (Fernandez-Albert, Llorach, et al. 2014).
For each metabolite, HMDB contains information including spectroscopic properties, relevant metabolite physiology, related enzymes and transporters, origin of the metabolite, known associations of metabolites, and disease phenotypes
Metabolites are identified in a hierarchical manner, by subdividing them from larger “kingdoms” into more specific “classes” and finer “families.”
A distinct advantage of metabolomics data is that the strength of association with disease outcome tends to be higher than what is observed with genetic data, where typical observed odds ratios are more modest
A metabolic phenotype, or metabotype, is the sum of the measured metabolites that exist in the biospecimen at a given time (Suhre
The metabotype can be influenced by a broad range of environmental exposures, from food intake, to hormonal changes, to the intake of medications.
The dynamic nature of metabolites enables the study of metabolic profile changes under varying conditions relevant to disease pathogenesis or treatment.
Clinical trials have also integrated metabolomics data to study whether metabolic profiles differ on and off medication.
Network medicine naturally applies to metabolic networks, whose nodes are separate metabolites that are linked together when they are involved in the same biochemical reactions
network theory suggests that the connections between the metabolites are not random, but rather follow a scale-free distribution
Gaussian graphical models (GGMs) are able to differentiate direct and indirect effects through the use of conditional dependencies, which result in partial correlations.
Bayesian networks are based on a data structure that encodes conditional probability distributions between variables of interest using a graph composed of nodes and directed edges
As such, metabo-profiles can be viewed as an intermediate measure that links predisposing genes and environmental exposures to a resulting disease state.
A study of twin pairs evaluated the longitudinal measures of plasma and urine metabolites obtained using NMR and found that many were heritable.
All of these -omics data provide information at different biological levels but all explain different aspects of the underlying systems biology.
Metabolite GWAS analyses (MGWAS) are typically performed using a regression model with the metabolite as the outcome, the single-nucleotide polymorphism (SNP) genotype as the predictor, and several covariates in the model for adjustment.
One of the first studies to integrate metabolites and epigenetics initially identified metabolites that were correlated with age and evaluated the association of this subset of metabolites with genome-wide epigenetic data.
Combined analyses of metabolite and gene expression data have great promise for identifying underlying regulatory networks that can link genes to the metabolome.
Two primary approaches have been described for integrating metabolomics and proteomics data
The first approach is knowledge-based and relies on known metabolic databases; the second approach uses the data to inform the functionality of the pathways.
The reference and draft networks are then compared and discrepancies between the two identified.
A second strategy for integrating multiple -omics data is to perform sequential pairwise comparisons between the available -omics types and to then link the results together.
Expression, SNP, and methylation probes with top associations to one or more metabolites were then included in a Bayesian network analysis. A conditional Gaussian Bayesian network (CGBN) was learned from this data using the CGBayesNets package in MATLAB version R2013b
This synergistic relationship between data and networks continues to extend from genomics, transcriptomics, and proteomics to newer regimes of high-throughput measurements, including metabolomics, epigenomics, and fluxomics.
Numerous publicly accessible repositories of -omics data have been amassed in recent years, which form the empirical foundations for network construction
Transcriptional regulatory networks (TRNs) associated with transcription factors have been constructed by integrating genomic sequence data with transcription factor binding information.
