The Trouble with Metabolomics

Within the area of systems biology, analytical scientists are most familiar with metabolomics because it uses well-known techniques (such as NMR spectroscopy, LC-MS, GC-MS) and because it is often swift to embrace new advances, such as high-resolution mass spectrometry (MS). But with multiple advanced tools comes much data. Indeed, a wealth of data points lay hidden in a given ion chromatogram, but some of these will be useless signals, so we need statistics to help us. And that’s where the tricky part begins.

Chemists need to be able to cope with concepts such as normalization, noise filtering, scaling, peak picking and alignment. And complex raw data is often treated with special algorithms that few people really understand. Multivariate statistics mine massive peak tables and reveal trends in the data, highlighting differences in physiology that are linked to metabolite concentration patterns. In the best-case scenario, markers are found – but these must be identified and validated, which is where a curtain often falls, obscuring the scene.

I truly believe metabolomics is the biggest challenge that the analytical community faces right now. Why? Because we must overcome problems that are not present or not as profound in other omics fields; for example, high diversity of molecular structures, huge differences in concentration ranges (several orders of magnitude), limitations in analytical response from different metabolites, a need for multiple analytical platforms (NMR/LC-MS/GC-MS, HILIC/IPLC/RPLC), ineffective data combinations, and research effort fragmentation and slow rate of metabolite identification. Then there are general issues that hinder expansion to the epidemiological level, including poor standardization in study design, sample collection/handling, chemical-statistical analysis and final reporting. And let’s not forget ineffective pathway analysis and fusion with other omics datasets.

To solve such complex problems, cross-disciplinary efforts are essential. Certain problems are largely analytically determined, so analytical chemists should take the lead in the quest for their resolution. A recent dispute shows the need for deep knowledge of analytical chemistry basics that, if left unchecked, cannot be corrected with increasingly sophisticated technology (1).
Therefore, we need to:

• standardize methods to develop common understanding
• improve the stability of our methods, by recognizing and overcoming analytical and pre-analytical issues that force methods to falter
• enforce quality control across different laboratories
• speed up and automate data mining in both targeted and untargeted modes, thereby improving the quality of the extracted information
• develop guidelines to help enforce accepted standards – from experiment design to data
  mining/deposition.
• run thousands of samples (including longitudinal samples) to generate trustworthy metabomaps of “controls” and solid data on the “normal” metabolite concentrations in different specimens
• establish public repositories to maximize the publication of “open-data”, which also needs to include spectroscopic data on unknowns and known-unknowns.

To promote standardization, there is a need for dedicated efforts from individual groups (2) and international coordination of multidisciplinary research (3). Such efforts should include training of researchers, the establishment of standard operating procedures (SOPs), and inter-laboratory testing.

SOPs may not be a perfect fit or tailor-made for each task, but I would happily use a debated, accepted, widely-applied procedure that provided me with data that  I could compare with others, even if I lost 20 percent of the metabolome coverage or sensitivity gained with my own bespoke method.

In the future, metabolomics may focus on a number of multi-analyte targeted and regulated methods and a smaller number of screening untargeted modes. Methods could be tailored to each specimen, assisting in diverse scientific fields, such as clinical decision making, food and drug safety, biotechnology, environmental monitoring and others. To achieve these goals, metabolomics needs major horizontal research grants and continuous endorsement from the analytical community. Such grants have supported the development of open-source free software (for example, MZmine) and open databases (for example, HMDB and LipidMaps), which have helped the metabolomics community immensely. Therefore, I call on international funding bodies to continue and strengthen their support of initiatives that present wide perspective and applicability.