By leveraging AI, Humanity’s biological age models can estimate biological age from any blood test panel.
We came across a new preprint published that has demonstrated (although still pending peer review) that an estimation of biological age can be made from almost any blood test panel.
A team from Humanity, the company behind an app that monitors your rate of aging and guides you on how to slow it down, and Imperial College London has used AI to leverage minimal data points available in any standard blood test panel to estimate biological age.
Their work shows that, if the full blood panel is measured, this platform renders an estimation of biological age with higher accuracy than Levine’s PhenoAge tests , and, excitingly, even if only a subset is available, meaningful results can still be obtained.
Longevity.Technology: Estimating biological age, whether of the whole person, individual organs or mapping against chronological age is big business. But results can vary wildly, with different tests producing different results, and having regular biological age tests can prove expensive. And which biomarker or combination thereof to use? Biological age has been estimated using a variety of biomarkers, including telomere length, DNA methylation, proteomics, metabolomics, glycomics and data gleaned from wearables – do we now need to look no further than any standard blood test?
The new study used the power of machine learning and a feature-set of 57 circulating blood biomarkers and all-cause mortality from 306,756 participants from the UK Biobank to improve the estimation of biological age. Participants’ ages ranged from 37 to 73 years .
The research team discovered that the developed Elastic-Net derived Cox model, which uses only 25 selected biomarkers, surpasses the widely-used blood-biomarker based PhenoAge model by an impressive 9.2% relative increase in predictive value .
The upshot is a practical and cost-efficient method of estimating biological age that can be achieved by using common clinical assay panels with only a few biomarkers, alongside imputing some markers which are not in the panel.. This breakthrough method offers a more accurate reflection of biological age that is easily scalable and can be used by the general population at large, looking at data from existing blood panels, as well as building blood tests into regular check-ups and other preventative strategies.
The new method will enable targeted interventions and preventative health strategies that not only promote healthy aging, but that will reduce the burden of age-related diseases and improve overall population health.
Understandably, we were intrigued by the preprint, so we approached Humanity for comment, speaking to co-founder Michael Geer.
He told us that he and Pete Ward founded Humanity to make a meaningful contribution to the longevity mission, leveraging their consumer tech background to take a variety of scientific tests and interventions and bring them to the mainstream.
“Many times unfortunately a lot of these great breakthroughs get left at proof of concept stage, as that is the important point to get to in science – publishing the results,” Geer explains. “However, unless someone then brings that knowledge directly to people in a way they can use, then it will not achieve the impact in society of which it is capable. It won’t reach its true potential.”
Geer explains that the publication released in preprint is another step in that direction and emphasized that this is still pending peer review and the team is looking forward to that robust process.
“This shows that we can use AI to not only increase the predictive power of Biological Age models, but more importantly, we can make them work with almost any combination of blood markers and thus applicable to nearly every blood test panel a person will get for almost any reason,” says Geer.
“When talking about measuring your Biological Age, it is really only useful if you can afford to do it frequently enough and it is available where you live. We have delivered digital marker Biological Age models, along with our partner Gero, now to every phone, and this published model we have released in preprint is the next big step to bring blood biomarker models in that same direction of availability to everyone.”