New study uses machine learning models for biological life prediction – with good accuracy.
Age is just a piece of paper – chronological age, that is. When it comes to biological age, however, there is a lot more information to be uncovered, from the age of individual organs, to just how much lifespan is left on the clock.
Longevity.Technology: Biological aging is the gradual deterioration of functional characteristics. If we had a comprehensive set of metrics that could predict lifespan on an individual basis – based on everything from DNA deletions to walking speed – we could drive tailor-made treatments for age-related diseases.
A team led by David Sinclair, professor of genetics in the Blavatnik Institute at Harvard Medical School, has used frailty data to power two AI-based clocks, which can gauge both chronological and biological age in murine models.
In a paper published in Nature Communications, the research team said the work, which tracked the health of 60 aging male mice for over a year, until they died naturally, is the first time that a study has tracked frailty for the whole length of a mouse’s life.
Using a set of non-invasive tests (including hearing and vision loss, spine curvature and gait) that generated an overall frailty score, the team developed two computer models to interrogate the data. The FRIGHT clock (Frailty Inferred Geriatric Health Timeline) predicts how biologically old a mouse is based on frailty score and the AFRAID clock (Analysis of Frailty and Death) predicts how much lifespan an old mouse has left, and can do so up to one year ahead. Predictions generated by the study had an accuracy to within two months.
“We are working to predict mouse health spans so we can quickly assess the effectiveness of interventions intended to extend life and move toward doing the same one day in humans,” said Professor Sinclair, senior author of the study .
To develop their research, the Harvard team tracked frailty in two sets of mice that had been given diets or therapy that had previously demonstrated life- or health-extending capabilities in earlier mouse studies. The aging clocks were able to accurately predict whether the interventions would make an improvement to biological age, or lead to longer life. The lab is also now expanding its studies to incorporate female mice.
“We want to understand how the aging process itself works so we can find ways to reduce the incidence of all these diseases together, rather than one at a time.”
“It can take up to three years to complete a longevity study in mice to see if a particular drug or diet slows the aging process,” said co-first author Alice Kane, Harvard Medical School research fellow in genetics. “Predictive biometrics can accelerate such research by indicating whether an intervention is likely to work.
“Diseases like cardiovascular disease, cancer, diabetes, neurodegeneration and even severe COVID-19 predominantly affect older people. We want to understand how the aging process itself works so we can find ways to reduce the incidence of all these diseases together, rather than one at a time .”
Michael Schultz, also a Harvard Medical School research fellow in genetics, says that given the correct data set, “it would be relatively straightforward” to develop a similar life expectancy clock for humans.
“However, such a large, longitudinal dataset that tracks people from their 60s into their 90s with significant mortality follow-up data is not available, to our knowledge,” he explained. “Finding a less expensive or less invasive way to test the molecular underpinnings of physiological signs like gait would help us make earlier and more accurate health span predictions and interventions .”
Humans are obviously more complicated, but the Harvard research is a positive step towards developing aging clocks that could help delay or avoid life-limiting diseases or conditions.