for clinicians

The genome is the genetic fingerprint. It is fixed at conception and is the same at eight years old as at eighty. It describes predisposition.

The metabolome is the biological fingerprint. It is the complete set of small molecules produced by metabolism: lipids and lipoprotein particles, amino acids, fatty acids, ketone bodies, glycolysis intermediates and inflammation markers. It describes the present state of the organism, and it changes with weight, diet, alcohol, sleep, exercise, medication and disease.

That difference matters clinically in two ways. A genetic result cannot be acted on and re-measured; a metabolic profile can. And a genetic result describes risk conferred at birth, whereas the metabolome carries the accumulated effect of everything since, which is where most modifiable cardiometabolic risk actually sits.

Every tissue and biofluid has a metabolome. Longitude measures the circulating metabolome, in plasma or serum, because blood perfuses every organ and therefore carries an integrated signal from the whole body rather than from one compartment.

A biobank is an organised repository of human biological material held for research, collected with consent. Not only blood: plasma, serum, DNA, urine and solid tissue of various types, stored under controlled conditions, often for decades.

The UK Biobank is the largest in the world. It recruited approximately 500,000 healthy volunteers aged 40 to 69 between 2006 and 2010. Each sample is linked to the respective donor’s NHS record, so hospital admissions, cancer registrations and deaths accrue against each participant prospectively. The cohort now carries fifteen to twenty years of hard outcome data attached to blood drawn before any of those events occurred.

That is the precondition for prediction. You need the baseline sample, and you need to know what happened next.

The models are trained and validated across more than a million participants in at least fourteen international biobanks, with UK Biobank, the Estonian Biobank and the Finnish biobank network as the core. More than 900 peer-reviewed papers have been published using the platform.

Cardiovascular disease and type 2 diabetes: as good, but much easier.

Against QRISK3 and QDiabetes the metabolomic scores perform comparably. However, QRISK3 requires age, sex, ethnicity, smoking status, diabetes and family history, a current blood pressure, a treated-hypertension flag, a BMI and a cholesterol ratio, with the appointment, measurement and data entry that implies.

The metabolomic score requires a blood sample, age and gender – nothing else. No family, medical or social history, no examination, no repeat visit, no assumption that the record is complete. For a population-scale programme that difference is the whole difference. Not because the prediction is better, but because it can actually be delivered to everybody.

Health economic modelling in Finland supports this. A metabolomic health check compared against the current FINRISK and FINDRISC standard was dominant in the base-case scenarios, more health at lower cost, and released substantial clinical capacity.

Chronic kidney disease and fatty liver disease: where there is no comparator.

These are the distinctive outputs, because no established primary care risk tool exists for either.

Chronic kidney disease is currently identified by eGFR, serum creatinine and urinalysis, a measure of function already lost rather than a forward risk estimate. Fatty liver disease is flagged by liver enzymes, which are insensitive and non-specific for steatosis. The metabolomic models outperform both, and do so prospectively.

Both conditions are common, silent until late, and substantially modifiable when found early. Neither is systematically screened for in general practice.

Because the metabolome is modifiable, the natural shape of the consultation is: what the profile shows, which contributing markers are movable, what intervention is indicated, and when to re-test to see whether it worked.

Every partner clinician is trained before their first patient, and we provide the AH Clinical Interpretation and Action Guide covering risk categorisation, recommended action at each level, and the common consultation scenarios.

A fair question from any clinician working in a diverse community is whether a model built largely in a European cohort transfers.

Two features make this platform more transferable than most. The biomarkers are reported in absolute concentration units rather than cohort-normalised scores, so the same coefficients apply in a new population without recalibration. And ethnicity is not an input variable. QRISK3 carries explicit ethnicity coefficients precisely because the clinical inputs it uses do not capture the differences. The working principle here is the opposite: population differences in cardiometabolic risk are themselves expressed in the metabolome, so they are already in the measurement.

Nightingale NMR data have been generated, and cohort analyses carried out, across European, East Asian, South Asian, North American and sub-Saharan African populations, including Biobank Japan, the Hong Kong cohorts, the Multi-Ethnic Cohort Singapore, the South Asia Biobank, the Bangladesh BELIEVE Study, the Mass General Brigham Biobank and Malawi Healthy Lives. These deployments and analyses have demonstrated cross-racial applicability of the risk stratification.

The honest caveat. No published validation cohort composed specifically of Afro-Caribbean populations, at anything like UK Biobank scale, exists. That is a recognised gap in the field rather than one specific to this platform: it affects every cardiometabolic risk prediction tool in use, and QRISK3 is known to perform poorly in Afro-Caribbean populations despite carrying an ethnicity term.

Our position is straightforward. The structural argument above, and the breadth of the existing base, give good reason to expect the stratification to hold. We say expect rather than know, because that is the state of the evidence. Deployment in Bermuda and the Caribbean is an opportunity to help close that gap rather than inherit it.