Using Metabolomics in COVID-19 research: COVIDIME study 

In an online interview, Dr Tim Dierckx from the University of Leuven, shares details about the findings of his latest research — the COrona VIrus DIsease Metabolome, or, COVIDIME study — and why he chose Nightingale’s NMR analysis to study the samples.

Many COVID-19 researches are currently trying to understand the disease through the genome lens. So, when Dr Tim Dierckx and his colleagues from the Rega Institute for Medical Research, University of Leuven, Belgium decided to study the development and severity of the infection through the metabolomics lens, it was uncommon.   

In an online interview with Nightingale, Dr Dierckx shares:  

  • The cohort study’s aims and how the findings will benefit both patients and healthcare providers,  
  • Why they chose Nightingale’s NMR panel and its advantages,  
  • Why GlycA, an inflammation biomarker in the panel, is interesting,  
  • The role metabolomics can play in other COVID studies,  
  • How Nightingale’s metabolomics data on 100,000+ samples from the UKBiobank can be a treasure-trove for the scientific community.   

His full interview:

Could you give us an overview of your COVID project — what the aim of the study is? Who is involved and your role in it?   

Dierckx: The current COVID-19 project is a collaboration between three different Belgian hospitals — Jessa Hospital, East Limburg Hospital (ZOL) and the University Hospitals of Leuven.    

We’re performing a retrospective study to identify a metabolic profile in patients’ blood that associates with the severity of the disease at the time of hospitalisation as well during follow-ups. We are trying to identify metabolomic differences between COVID-infected patients with severe disease and those with mild disease and testing if any of the metabolic markers associates to the disease duration and mortality. Differentiating between severe and mild cases helps high-risk patients in getting better-targeted care, but it also helps medical providers to more efficiently manage their capacity and resources.   

Our research team at the Rega Institute for Medical Research at the University of Leuven is a relatively small team at the virology laboratory. This means we play many different roles — from sample handling to data analysis — but this has the advantage of being able to adapt to changes quickly and being able to move fast. In relatively short order, we’ve managed to have over 1,000 COVID-19 patient samples analysed using Nightingale’s platform.     

Many COVID studies have a genetics and gene-expression focus. What made you investigate metabolomics for your research?    

Dierckx: Many of my immunologist and virologist colleagues are currently studying COVID-19 from the gene-expression point-of-view. But we chose metabolomics because it offers a view that’s a step closer to reality than transcriptomics is.    

People often use the analogy “your DNA is a cookbook” — and the outcome of the recipes, the eventual outcome of your genes, your body, is made of protein. So, if one wants to study reality, they should ideally investigate the proteome, but this is expensive and hard to do. So, gene transcript analysis, transcriptomics, is usually performed because it’s the best way to get a comprehensive snapshot of a person’s gene activity, which is only a single step removed from the protein reality.  

This is where metabolomics can provide a compromise: it sits somewhere between the transcriptomic gene-expression snapshot and the proteomic reality. It doesn’t result in a complete and comprehensive picture like transcriptomics would, and the proteome still provides much more detail. But NMR metabolomics is a very cost-effective and straightforward way of capturing many aspects relevant to a person’s health. And perhaps the main reason we chose metabolomics is because there are some biomarkers in the NMR panel, such as GlycA, which you can’t measure with other techniques. All this made metabolomics the right choice for us.   

Are there any specific biomarkers in the NMR panel that are especially interesting to you? If yes, which are those and what makes them stand out?    

Dierckx: The main reason behind why NMR metabolomics caught my attention is Glycoprotein Acetylation, GlycA. NMR is the only platform to quantify this unique biomarker. For us, GlycA is very interesting as it’s a very good biomarker to study inflammation — it outperforms a lot of conventional inflammatory biomarkers, including C-reactive protein (CRP). GlycA seems to capture baseline inflammation much better than CRP. CRP being a single molecule is highly volatile — goes up and down quickly and sometimes isn’t present even though you’d expect it to be.  GlycA, on the other hand, is a lot more stable as it's a composite biomarker made up of several molecules. This means that in contrast to CRP, GlycA has baseline levels. So, you can see when it goes up, but also when it goes down, meaning you can identify anti-inflammatory states — something CRP can’t do.    

We observed this difference in our research as well when studying inflammatory bowel disease. Inflammation is the hallmark of these diseases, it’s literally in the name, and yet CRP wasn’t increased at all in a quarter of the patients with active disease. But when we checked for GlycA, we found they do have elevated GlycA concentrations.   

We use the term inflammation as if it’s this single thing. But there are many different causes of inflammation, and there are different types of inflammation that can result from each of these causes. GlycA seems to be affected by all kinds of inflammation — making it a perfect biomarker for baseline information on inflammation.    

The best part about the NMR metabolomics analysis is that while there may be one crucial biomarker we’re interested in (in our case, GlycA), it is accompanied by 200 other biomarkers which are a very rich source of complementary information. For instance, we are also looking at lipoprotein particles, fatty acids, amino acids and many other such options that research shows have a strong correlation with inflammation. And in larger cohort settings where the goal is to predict outcomes, a combination of biomarkers usually outperforms models based on single measurements. 

Nightingale’s recent UKBiobank-based NMR study has found a metabolic fingerprint that can predict the risk of severe COVID-19 years before the onset of the disease. How is the study different from your research setup?    

Dierckx: While Nightingale’s research identified a metabolic fingerprint for the risk of getting a severe COVID-19 infection in a healthy person, our research is looking for the metabolic fingerprint of that severe infection. We want to find out what makes a severe COVID infection look different from a mild COVID infection, on a metabolic level. 

Nightingale’s is a population-based study that took every single sample from the UKBiobank (taken years before infection) and then compared it to their current clinical data to find a common fingerprint among those severely infected by the virus. 

Ours is a cohort study, where every single one of the 1000+ samples comes from COVID-infected patients, and these samples were taken at hospital admission. What we study is how severe COVID-19 infection affects a patient and we’re trying to find if there are metabolic markers that can be used to predict how long an infected person will be hospitalised for and how likely they are to survive the disease. 

Do you think metabolomics can play a more prominent role in COVID studies? What could be some other potential use of metabolomics in COVID research?    

Dierckx: I feel, not just COVID research, but almost all clinical research can benefit from the inclusion of metabolomics. It’s easy-to-do and economical, but that aside, it provides a very good picture of the overall health of a person — and that’s an essential aspect in any given clinical research.   

Going back to COVID, think about vaccine research or a clinical trial for COVID-19 treatment. The baseline health of the participants plays a large role in these settings. If you want to profile each of your patients, you’ll typically need access to all their clinical data, and if you want to characterize the state of their immune system then you’ll need to do many kinds of protein tests, etc. etc. Using metabolomics, by doing just one analysis, you already get a very good overview of the patients’ baseline health. 

Before the COVID-19 epidemic, you have been involved in studies on various infectious diseases, but also systemic lupus erythematosus and inflammatory bowel disease. Which other areas of disease research do you think could also benefit from the metabolic analysis? Could you share a few examples?    

Dierckx:  Because the general health of the patients is essential for just about all disease research, I think metabolomics can be relevant to almost all medical research questions. From my own work, I’ve certainly seen evidence that suggests that the baseline inflammation levels, summarised by GlycA, play an important role in many more disease contexts than is currently suspected. 

We’ve shown GlycA’s use in inflammatory diseases like Lupus and are still exploring its relevance further in those contexts. For my post-doctorate study, I was also looking at parasitic and viral infections (and that’s why it was easy for me to switch to COVID). But then there’s the relationship between inflammation and ageing — studies concerning healthy ageing and frailty. Or cholesterol and inflammation — the NMR has an incredibly detailed panel (100+ measures) covering the lipoproteins. 

Just throw a dart in the dark, and you’re likely to hit a topic where metabolomics can be applied. 

Nightingale has analysed over 120,000 blood samples from the UKBiobank, and the metabolomics data will soon be available for the whole scientific community. What potential do you see in such a resource?  What possibilities does it open for the researchers?   

Dierckx: I don’t think there’s any other bigger single source/collection of metabolomic data, or information on public health, available another than the one via UKBiobank. So, the potential is enormous. It’ll be interesting to see how much of clinical data is released alongside it because that’ll determine how influential and impactful the research can be in that setting. Even pure metabolomics data (without demographics and clinical information) is still going to be impactful — it’ll act as a useful reference population. But adding that extra information would have tremendous value. As a bioinformatician, I’ll be surprised if people won’t use machine learning to dissect the entire data pool for everything it’s worth — there’ll be tons of opportunity there.