From correlation to causation: analysis of metabolomics data using systems biology approaches

Abstract

IntroductionMetabolomics is a well-established tool in systems biology, especially in the top–down approach. Metabolomics experiments often results in discovery studies that provide intriguing biological hypotheses but rarely offer mechanistic explanation of such findings. In this light, the interpretation of metabolomics data can be boosted by deploying systems biology approaches.ObjectivesThis review aims to provide an overview of systems biology approaches that are relevant to metabolomics and to discuss some successful applications of these methods.MethodsWe review the most recent applications of systems biology tools in the field of metabolomics, such as network inference and analysis, metabolic modelling and pathways analysis.ResultsWe offer an ample overview of systems biology tools that can be applied to address metabolomics problems. The characteristics and application results of these tools are discussed also in a comparative manner.ConclusionsSystems biology-enhanced analysis of metabolomics data can provide insights into the molecular mechanisms originating the observed metabolic profiles and enhance the scientific impact of metabolomics studies.

Pharmacomicrobiomics: The Holy Grail to Variability in Drug Response?

“…Identified metabolic biomarkers of response can be used to predict an individual's drug response. The metabolite content of the human body, and its ability to alter drug response, is dependent on the individual's genome, immune system, microbiome, diet, lifestyle, and external environment . Hence, to accurately predict drug response, it has become increasingly important to shift from scattered individual observations to an integrative, systematic approach that considers multiple co variates from different fields.…”

“…The metabolite content of the human body, and its ability to alter drug response, is dependent on the individual's genome, immune system, microbiome, diet, lifestyle, and external environment. 19,23,24 Hence, to accurately predict drug response, it has become increasingly important to shift from scattered individual observations to an integrative, systematic approach that considers multiple co variates from different fields. Although very few studies have used this integrative approach for drug response predictions, a report in a healthy, pre diabetic Israeli cohort showed that glycemic response can be accurately predicted for different realmeals using subject-associated variables, including blood hemoglobin a1c levels, diet, lifestyle, physical activity, and feces-associated microbial metagenomics data, 25,26 which will also be discussed in detail below.…”

Corruption of the Pearson correlation coefficient by measurement error: estimation, bias, and correction under different error models

“…To show the effect of additive uncorrelated measurement noise we consider the concentration profiles of three hypothetical metabolites P1, P2 and P3 simulated using a simple dynamic model (see Figure 2A and Section 6.1.2) where additive uncorrelated and correlated measurement noise is added before calculating the pairwise correlations among P1, P2 and P3: also in this case the magnitude of the correlation is attenuated, and the attenuation increases with the error variance (see Figure 2B). This has serious repercussions when correlations are used for the definition of association networks, as commonly done in systems biology and functional genomics 10,23 : measurement error drives correlation towards zero and this impacts network reconstruction. If a threshold of 0.6 is imposed to discriminate between correlated and non correlated variables as usually done in metabolomics 22 , an error variance of around 15% (see Figure 2B, point where the correlation crosses the threshold) of the biological variation will attenuate the correlation to the point that metabolites will be deemed not to be associated even if they are biologically correlated leading to very different metabolite association networks (see Figure 2C.…”

PhenoMeNal: Processing and analysis of Metabolomics data in the Cloud

“…MS and NMR data processing usually involves selection of parameters (which are often specific to the analytical instrumentation), algorithmic peak detection, peak alignment and grouping, annotation of putative compounds and extensive statistical analyses [6,7]. Many open source tools have been developed that address these different steps in data processing and analysis.…”