In our previous blog post, we discussed how large-scale biobank cohorts are generating massive metabolomics and proteomics datasets. 

Since then, Nightingale Health has announced two large projects in the UK Biobank: expanding the range of metabolites quantified and generating a new brain-related proteomics dataset to study neurodegenerative diseases using Alamar’s high-sensitivity platform. 

Today, we’d like to shift focus and review some recent multiomic research into cardiovascular disease (CVD).

 

The metabolomic profile of cardiovascular disease

The molecular epidemiology of CVD has been a core research focus at Nightingale since before our founding, going back to the study by Würtz et al. (2011) of subclinical carotid intima-media thickness. Today, our NMR CoreMetabolomics platform extends well beyond a standard lipid panel to quantify particle size, concentration and composition. Fatty acids, amino acids and other metabolites fill out the profile of cardiovascular health.

Fall et al. (2025) investigated how these measurements relate to atherosclerosis in the Swedish CArdioPulmonary bioImage Study (SCAPIS). In nearly 5,000 participants, they found that the associations of certain lipoprotein measures differed between coronary and carotid atherosclerosis. For instance, each doubling of intermediate-density lipoprotein concentration was associated with 26% higher odds of a raised coronary artery calcium score, but 124% higher odds for more extensive carotid plaques. 

Morze et al. (2025) used UK Biobank metabolomic data to investigate how lipoprotein particle parameters can predict future coronary artery disease (CAD). They found that most of the association between CAD and VLDL, IDL and LDL particle subclasses was accounted for by ApoB-containing particle number. The size and type of these particles were less important than particle number, which is directly measured by NMR.

 

Adding proteomics to the mix

The SCAPIS study also generated targeted proteomics data alongside metabolomics. In contrast to the lipid measures, the proteins showed broadly similar profiles of association with carotid and coronary atherosclerosis. Shared signals included the growth factor SCF, the anticoagulant TFPI and the cytokine decoy receptor osteoprotegerin. Adjusting for lipid parameters suggested that some proteins tracked lipid levels (e.g. PCSK9) and others had independent associations (e.g. SCF). 

The two technologies can also be used to study the same pathway. Corianò et al. (2026) investigated inflammation using Nightingale and Olink data in the UK Biobank. The NMR-measured GlycA gives a general measure of systemic inflammation, whereas individual cytokines and growth factors measured by Olink give insight into more specific signalling pathways. They confirmed the association between GlycA and major adverse cardiovascular events (MACE) and showed associations with MRI measures of cardiac structure and function. However, they found evidence that these relationships were most strongly statistically mediated by different inflammatory proteins (IL-6 for MACE, IL-1RA for left ventricular end-diastolic volume), suggesting that different inflammatory pathways could be involved.

 

Predicting different CVD endpoints with multiomics

Luo et al. (2026) developed neural-network-based scores from UK Biobank metabolomic and proteomic data to predict six cardiovascular outcomes: coronary artery disease, stroke, heart failure, atrial fibrillation, peripheral artery disease and venous thromboembolism. They found that both metabolomics and proteomics significantly improved prediction beyond conventional clinical predictors, and combining both molecular scores improved prediction beyond either score alone for CAD and stroke. Interestingly, metabolomic signals tended to be shared across diseases, whereas proteomic signatures tended to be more disease-specific. 

Du et al. (2026) investigated a similar question across 17 diseases, including five cardiovascular outcomes, focusing on how predictive performance depended on available clinical data. For four of the outcomes, comprehensive clinical data (including diagnoses, physical measurements and clinical blood tests) plus proteomics performed similarly to the full multiomic model, with metabolomics providing significant additional improvement only for peripheral artery disease. However, with only age and sex as clinical predictors, metabolomics was much more helpful; adding metabolomics on top of proteomics improved prediction for three of the five endpoints.

 

The complementary strengths of multiomics

In some cases, the two data types pull apart different pathways, such as the distinct metabolomic and proteomic signatures found by the SCAPIS study. However, multiomics can also give different views of the same underlying pathway, as in Corianò et al.’s finding that different inflammatory proteins statistically mediated GlycA’s associations with different cardiovascular traits. 

There is a recurring pattern in these studies of metabolomics providing more generalised biomarkers across cardiovascular traits, with proteomics picking out more disease-specific signatures. But even this is not universal: SCAPIS found different lipoprotein associations with coronary and carotid atherosclerosis, while proteomics told a story of mostly shared associations.

 

Expanding the opportunities for cardiovascular research

Nightingale Health offers NMR-based CoreMetabolomics alongside proteomics services using Olink and Alamar technologies, allowing researchers to generate complementary molecular measurements from the same samples. Researchers can choose targeted protein panels for specific pathways or broader platforms for more exploratory research.

CoreMetabolomics is also expanding, adding further opportunities for cardiovascular research. CoreMetabolomics already measures broad fatty acid classes, including omega-3 and omega-6. The expanded panel adds new measures including EPA, DPA and alpha-linolenic acid within omega-3, arachidonic acid within omega-6, and omega-5, omega-7 and omega-9 fatty acids.

Our analyses in UK Biobank illustrate why this additional resolution can matter. While all measured omega-3s were associated with lower stroke risk, the pattern was not consistent across individual subtypes for heart attack and overall cardiovascular disease. This created opportunities to investigate questions such as whether different omega-3 fatty acids have distinct relationships with cardiovascular outcomes.

CoreMetabolomics is also expanding its cardiovascular coverage with an NMR-derived Lp(a) prediction score. The score identifies individuals likely to have high Lp(a), providing a way to investigate Lp(a)-associated cardiovascular risk at scale without an additional assay. Unlike directly quantified biomarkers in the panel, the Lp(a) output is a prediction score rather than an absolute measurement.

As datasets grow and assay throughput increases, new opportunities are opening up to do truly population-scale multiomic science. Combining these assays offers new ways to investigate cardiovascular disease biology.

Interested in combining metabolomics and proteomics in your cardiovascular research? Contact the Nightingale Health research team at research@nightingalehealth.com to discuss your study.

By Luke Jostins-Dean, Principal Scientist at Nightingale Health

 

References

  1. Würtz P, Soininen P, Kangas AJ, et al. (2011). Characterization of systemic metabolic phenotypes associated with subclinical atherosclerosis. Molecular BioSystems, 7(2), 385–393. https://doi.org/10.1039/C0MB00066C
  2. Fall T, Gummesson A, Hammar U, et al. (2025). Plasma proteomic and metabolomic profiling of coronary and carotid atherosclerosis in the SCAPIS study—differences and similarities. Cardiovascular Research, 121(11), 1649–1652. https://doi.org/10.1093/cvr/cvaf076
  3. Morze J, Melloni GEM, Wittenbecher C, et al. (2025). ApoB-containing lipoproteins: count, type, size, and risk of coronary artery disease. European Heart Journal, 46(27), 2691–2701. https://doi.org/10.1093/eurheartj/ehaf207
  4. Corianò M, Tahasildar S, Huang L, et al. (2026). Gene-environment interactions shape cytokine-mediated inflammation and cardiovascular risk. European Journal of Preventive Cardiology. Advance online publication. https://doi.org/10.1093/eurjpc/zwag435
  5. Luo Y, Zhang N, Yang J, et al. (2026). AI-based multiomics profiling reveals complementary omics contributions to personalized prediction of cardiovascular disease. Nature Communications, 17, 2269. https://doi.org/10.1038/s41467-026-68956-6
  6. Du J, Zhou M, Wang H, et al. (2026). Multi-omics integration predicts the incidence of 17 diseases in the UK Biobank. Nature Communications, 17, 6271. https://doi.org/10.1038/s41467-026-73017-z