Researchers have performed an in-depth study on artificial intelligence-driven aging clocks that assess health and lifespan based on blood data.
Researchers at the Institute of Psychiatry, Psychology & Neuroscience (IoPPN) at King’s College London have carried out an extensive study to assess artificial intelligence-based aging clocks, which estimate health and lifespan using information from blood samples.
In their research, the team trained and tested 17 machine learning algorithms using blood marker data from over 225,000 participants in the UK Biobank, aged between 40 to 69 years at the time of their recruitment. They explored how effectively various metabolomic aging clocks predict lifespan and how strongly these clocks correlated with indicators of health and aging.
A person’s metabolomic age, referred to as “MileAge,” reflects the internal aging of their body based on blood metabolites. Metabolites are tiny molecules generated during metabolism, such as when the body converts food into energy. The difference between a person’s metabolite-predicted age and their actual age, called MileAge delta, shows whether their biological aging is speeding up or slowing down.
This study, published in Science Advances, is the first to extensively compare machine learning algorithms regarding their effectiveness at creating biological aging clocks using metabolite data, utilizing one of the world’s largest datasets. It received funding from the National Institute for Health and Care Research (NIHR) Maudsley Biomedical Research Centre (BRC) and drew from UK Biobank data.
Participants whose biological aging was accelerated (i.e., those with a metabolite-predicted age greater than their chronological age) tended to be frailer on average, more prone to chronic illnesses, rated their health negatively, and faced a higher risk of mortality. They also exhibited shorter telomeres, which are protective structures at the end of chromosomes associated with aging and age-related diseases like atherosclerosis. In contrast, those with reduced biological aging (having a metabolite-predicted age younger than their chronological age) showed only weak correlations with better health outcomes.
Aging clocks have the potential to identify early indications of health decline, facilitating preventive measures and interventions before diseases arise. They may empower individuals to monitor their health actively, make healthier lifestyle decisions, and take steps to maintain their wellness for a longer duration.
Dr. Julian Mutz, King’s Prize Research Fellow at the IoPPN and the study’s lead author, stated: “Metabolomic aging clocks could provide valuable insights into identifying individuals who may be at higher risk for health issues later on. Unlike chronological age, which is fixed, our biological age can be adjusted. These clocks serve as an important proxy measure for biological age in biomedical and health research, aiding individuals in making informed lifestyle choices and guiding preventative strategies by health services. Our study assessed a wide range of machine learning techniques for developing aging clocks, revealing that non-linear algorithms are the most effective at capturing aging signals.”
Professor Cathryn Lewis, Professor of Genetic Epidemiology & Statistics and Co-Deputy Lead of the Trials, Genomics and Prediction theme at the NIHR Maudsley BRC, and a senior author of the study, remarked: “There is considerable interest in creating aging clocks that accurately evaluate our biological age. Advanced big data analytics are key in developing these tools. This research represents a significant advancement in recognizing the potential of biological aging clocks and their capability to guide health decisions.”
The researchers discovered that a metabolomic clock produced using a specific machine learning algorithm known as Cubist rule-based regression showed the strongest association with various health and aging indicators. They also found that algorithms capable of modeling non-linear relationships between metabolites and age tended to perform best at identifying biological signals relevant to health and lifespan.
