Protein Signature Linked to Death Risk Could Aid Life Extension Efforts

By Deborah Borfitz 

December 16, 2025 | A biological signal of death can be found in the blood five to 10 years in advance of that mortality outcome, a fascinating and unexpected discovery made in the quest for a signature marker of multi-morbidity by researchers at the University of Surry (UK). While by no means a clinically validated predictive test, the protein panels hold promise in motivating lifestyle changes and food choices that could change that poor health trajectory, according to Nophar Geifman, Ph.D., professor of health and biomedical informatics.  

Regardless of the ultimate cause of death, or how the data was sliced, five key proteins (PLAUR, SERPINA3, CRIM1, DDR1 and LTBP2) “seemed to constantly pop up,” she says in reference to the study that was published recently in PLOS One (DOI: 10.1371/journal.pone.0336845). Based on the known associations of the proteins with inflammation and cancerous-like activities, it made sense that they might be related to increased risk. 

The study was based on an analysis of the blood protein profiles of over 38,000 predominantly white British people in the UK Biobank study who were recruited 15 to 19 years ago, focusing on adults aged 40 to 69. The objective was to determine if their risk of early non-accidental mortality was reflected by levels of certain circulating proteins in their blood.  

With further studies on completely independent and diverse cohorts, ensuring adequate precision, the findings could be turned into a clinical test for identifying individuals most at risk of mortality to prompt earlier medical intervention, Geifman says. Health self-monitoring, a cornerstone of behavior change, is growing rapidly and globally, as evidenced by the popularity of direct-to-consumer gut microbiome tests and genomic sequencing services. “These are all thriving services ... especially in places where healthcare systems are either overwhelmed or not equally accessible to everyone.” 

Geifman says she sees the potential for a risk-of-dying test to aid countries like the UK that are investing in screening and preventive medicine.  

The “worried well” will undoubtedly be the most engaged population group, she adds. “There’s a huge section of the population that probably doesn’t want to know [their mortality risk] and will probably continue with behaviors that confer ill-health such as smoking, poor food choices, and not getting annual health checks. Those are always going to be the harder-to-target demographics.” 

Disease Connection 

This was but one of a series of scientific investigations Geifman and her colleagues have undertaken with the UK Biobank, an unparalleled biomedical resource holding genetic, lifestyle, and health data from half a million British adult volunteers. This includes proteomic and metabolomic measurements and genomic sequencing on the overall cohort as well as brain/heart/abdomen MRIs on a large subset of those individuals. 

The “extra bonus” is that the information is linked to the electronic record system of the National Health Service, meaning all conditions coded in that digitized system are also known, says Geifman. This has enabled various published studies over the past few years around how diet influences health, including one late last year finding that vegetarians who eat ultra-processed vegetarian meat alternatives (e.g., veggie burgers) suffer from depression more than vegetarians who do not (Food Frontiers, DOI: 10.1002/fft2.532).   

In aging populations, different co-occurring conditions—e.g., diabetes, hypertension, and dementia—are a big problem. The elevated burden of having multiple diseases, and complexity of managing them, is why the research team was interested in looking for a signature marker of multi-morbidity in the blood 10 years before those diseases start developing, she explains. 

Since there are dozens of chronic conditions that might occur in multiple combinations, they decided to, initially, simplify the search by focusing on the finite and easy-to-define outcome of mortality. Many chronic diseases are hard to pin down due to patient heterogeneity, disease subtypes and inaccurate diagnoses, as well as errors in the way diseases are coded in electronic health records, Geifman notes. Death is “almost always” accurately documented. 

They first tested each of the 3,000 or so circulating proteins measured in their study cohort to see how many were associated with an increased risk of dying within five or 10 years, she reports. Two kinds of statistical models were used for that analysis, one making basic adjustments for BMI, sex and age and another that considered different lifestyle factors such as whether participants smoked or exercised. 

This revealed 392 proteins that were associated to a statistically significant degree with increased five-year risk and 377 proteins associated with 10-year increased risk of all-cause mortality, excluding accidents.  

Next, researchers looked at whether the proteins were associated with disease-specific mortality and found different signatures for the main causes of death such as cardiovascular disease and cancer. Here, there was overlap with 19 proteins between them, all being somehow involved in disease development after adjustments were made for the known health risk factors, she continues. 

Finally, they used those proteins to build a predictive model and compared its performance alone and in combination with other death-associated metrics such as age, BMI, and sex. Six of the proteins were top hits for the five-year risk, and 10 proteins for the 10-year risk, with a five-protein overlap between them. “Those are the five hits that seemed to carry the most association with risk regardless of timeframe and whether you adjust for those other lifestyle factors,” says Geifman. 

“We can’t say there’s a causal relationship because we haven’t investigated that,” she adds. The directionality of the relationship—whether elevated levels of these proteins cause early mortality, or the risk of early mortality increases levels of these proteins—would in any case be difficult to establish. 

But the findings do open a world of research opportunities, mostly for other researchers to pursue, says Geifman. “We’re predominantly a dry lab, so while we carry out data-driven investigations using existing large data resources and sophisticated analytics such as machine learning, we don’t sit in the lab and run samples and measure biomolecules ourselves.” 

Modifiable Risk 

While there are no immediate plans to turn the protein signature into a clinically useful test, Geifman and her team routinely work with other researchers interested in developing diagnostic panels, she says. Those developed to date have tended to be “very disease-focused where there is an immediate clinical need.”  

The latest study is the first step in a “death markers project” now underway in the Geifman lab looking at how to incorporate the protein signature of mortality into the multi-morbidity aspect of their research, she adds. The question being addressed is whether those proteins levels can be modified by treating chronic conditions in new and different ways to change the course of disease and improve life expectancy. 

A lot of prior work that focused on food choice, including the study on vegetarians, was led by Hana Navratilova (now Ph.D.). This notably includes a 2024 paper using a machine learning approach to stratify people based on what they say they like to eat, says Geifman. It turns out that food preferences are associated with distinct health outcomes and metabolomic and proteomic profiles (Journal of Translational Medicine, DOI: 10.1186/s12967-024-05663-0). 

Over 180,000 individuals in the UK Biobank were statistically profiled to reveal three endotypes— a health-conscious group who prefer vegetables and fruits over animal-based or sweet foods; an “omnivore” group who like just about everything; and a “sweet-tooth” group who favor sugary desserts and beverages. In terms of differences in disease occurrence, healthy eaters had a lower risk of heart failure and chronic kidney disease, while the sweets lovers had greater risk of depression, diabetes, and stroke.  

In another 2024 paper, Navratilova and Geifman again tapped the UK Biobank to show a healthy dietary pattern can significantly reduce the risk of cardiovascular disease, colorectal cancer, and type 2 diabetes (Nutrients, DOI: 10.3390/nu16040523). More recently, they proposed a digital health intervention catering to individual needs and preferences to inspire dietary choices reducing cardiovascular disease risk (Journal of Internet Medical Research, DOI: 10.2196/75106). 

The newly discovered protein signature is in no way deterministic, Geifman stresses. Even people who are currently at high risk of a poor outcome, such as dying within five or 10 years, can do a lot to reduce that risk and postpone or avoid poor outcomes—beginning with the common sense, often-repeated health advice to exercise more, eat better, and get enough sleep. “People have a lot of control over their own health.”