A recent publication provides a new perspective on the “high fidelity” representation of the metabolome hidden within global metabolic profiling data.

Research from the Section of Bioanalytical Chemistry, in the Department of Metabolism, Digestion and Reproduction, aims to use HiFi (High Fidelity) Metabolomics to increase accurate biological interpretation of global data profiling and improve the translation of results into the clinic environment. We spoke with Dr. Caroline Sands about the work.

What is the HiFi metabolome?
In global profiling metabolomics we aim to capture measurements for as many metabolites as possible in a sample to provide the highest coverage. A really good technique for this is liquid chromatography−mass spectrometry (LC-MS), which combines the separating power of liquid chromatography (to separate the metabolites) with the detection of mass spectrometry (to create charged ions from the metabolites and then detect them). However, LC-MS produces complex datasets where each metabolite is represented by multiple measurements (commonly called features, for example, derived from different ions produced in the mass spectrometer as a metabolite is broken apart). We wanted to find out if there was any way of determining which features represent the underlying metabolite concentration with the highest fidelity, so we can constrain our final datasets to be the most representative of the truth.

Why is it important?
For many reasons! Perhaps foremost, to ensure accurate biological interpretation and the translation of results into the clinic, where reliability and performance are critical. But even before we get there, it is incredibly important in modeling, and especially practically, to save time wasted trying to determine the metabolic identity of a feature whose intensity measurements deviate from what is actually in the sample.

How will your findings impact future research?
We hope it will push forward-thinking in the field about how we can assess global profiling feature measurements in terms of relative accuracy. In targeted and bioanalysis approaches, it is standard to run a calibration curve for each metabolite (using chemical standards) to ensure measurement accuracy, we are promoting an analogous approach. Of course, we cannot run standard calibration curves for all metabolites in global profiling as we inherently don't know what we are measuring before we start! But we can do the next best thing, by using a set of differently diluted (but otherwise identical) samples to follow the response of every feature at different concentrations and create a metric of how well (or badly!) each feature measurement responds to increasing sample concentration.

How did this research come about?
While the conclusions drawn are anticipated to have a broad impact on the field of metabolomics, the line of thinking emerged from a single case of intra-departmental collaboration with colleagues Prof. Waljit Dhillo and Dr Chioma Izzi-Engbeaya (Department of Metabolism, Digestion and Reproduction). When reviewing their data and comparing published global profiles to subset measurements derived from a tool we were developing called PeakPantheR, we noticed some discrepancies between the ions from the original dataset and our new measurements. Depending on the type of ion, we observed different trends when we looked in the dilution series samples, and that's where it all started. Effectively, it is an excellent example of what can happen when clinicians, analytical chemists and bioinformaticians work together under one roof to dig deeper into complex data.

What are the next steps?
Putting all this theory into practice! Our perspective piece really advocates the theory, with some detailed examples but, as always, applying this in the real world will take some time. Watch this space! And if anyone reads our paper and has their own ideas, please we would love to hear from you.

Representing the Metabolome with High Fidelity: Range and Response as Quality Control Factors in LC-MS-Based Global Profiling:
Caroline J Sands, María Gómez-Romero, Gonçalo Correia, Elena Chekmeneva, Stephane Camuzeaux, Chioma Izzi-Engbeaya, Waljit S Dhillo, Zoltan Takats, Matthew R Lewis. PMID: 33448796 DOI: 10.1021/acs.analchem.0c03848