AI-mediated understanding of biological age—as opposed to chronological age

Youth is a prized asset in our society—perhaps the most prized asset of all. Understanding its inherent value, younger people strive to exploit their youth; older people, meanwhile, work to “recapture” it. Obviously, what they try to recapture is not youth per se but the key attributes of youth, including smooth skin, toned muscles, thick and shiny hair, good health, sexual vigor, an absence of chronic pain, high energy and clarity of thought. As each of us comes to realize, we have no control over our chronological age; time moves on regardless of how we feel about the fact. We can, however, do things to freeze or roll back our biological age—our “true” age, the number much more closely associated with health status and mortality than our chronological age. A healthy diet, regular exercise, proper sleep, good hygiene, and a strong relationship with friends and family have all been shown to slow chronological aging and promote greater longevity. A 50-year old whose biological age is closer to 35 than 50 is on the path to a long and healthy life. It is biological age, in short—as revealed by key “bio-markers”—that tends to correlate most closely with how we feel and how long we live. As a result, accurately assessing these bio-markers is crucial to predicting future health status and longevity and—if necessary—organizing lifestyle changes and/or medical interventions.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                           

The fundamental problem, until quite recently, is that researchers had very little success identifying reliable measures of biological age. Now, however, researchers are having much greater success identifying a number of reliable biomarkers that take into account key behavioral, environmental and genetic factors. Everything from telomere length and T-cell counts to DNA methylation levels have been shown to correlate with health status and mortality—and thus an individual's true biological age. While immensely valuable, however, these biomarkers have tended to lack precision. They've been incredibly helpful—but not as helpful as they might be.

The application of AI and data analytics has changed everything. AI-based systems are now able to create and analyze bio-marker datasets much larger than anything possible previously. Both the amount of data and the number of potential biomarkers is expanding rapidly. The Aging Analytics Agency (AAA), a UK-based analytics firm focused heavily on longevity research, relies on data drawn from at least fifty distinct biomarkers to produce its findings. Each biomarker is associated with the development of specific age-related diseases and syndromes. AAA, and startups like Elysium Health, Chronomics, and Cambridge Epigenetix (also UK-based) are dedicated to creating and marketing highly-accurate epigenetic clocks— “aging clocks,” in essence—that analyze biomarkers within DNA. And since AI produces increasingly reliable results in response to increasing amounts of data, the application of highly sophisticated algorithms to more and more bio-marker data cannot help but reveal additional reliable biomarkers over time. This is important, as there seems to be growing consensus that accurately assessing biological health and longevity requires researchers to pay close attention to multiple bio-markers—not just one or two.                                                                                                                                 

AI's enormous contribution to understanding biomarkers and “aging clocks” was recently highlighted by Insilico Medicine, a biotech company based in Johns Hopkins University’s Emerging Technology Centers. Insilico blends genomics, big data analysis, and AI to help discover new drugs. Alex Zhavoronkov, lead author of a landmark 2019 study and Insilico's CEO, suggests that while most of us are good at assessing the age of our fellow human beings using a combination of sight, sound, touch, and even smell, deep neural networks (algorithms designed to mimic the processes of the human brain) do even better. Insilico's neural networks draw on a complex web of biomarkers developed by the company to assess an individual's biological age and potential longevity. The results have been promising. Pharmaceutical companies, insurance companies, biotech companies, and a wide array of actors with an interest in longevity are all keen to exploit these advances. Polina Mamoshina, Insilico's Senior Scientist and a fellow author on the study, claims that advanced understanding of deep biological aging clocks will soon prove essential to biological target identification—that is, identification of the biological entities (proteins, nucleic acid, genes, etc.) whose behavior is affected by compounds that companies hope to develop into pharmaceutical drugs. In other words, AI is helping drug companies make sense of vast amounts of bio-marker data to gain a clearer picture of how substances act within the human body. Other important potential applications of this research include assessing the value of collected data and data quality control.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                          

One especially fascinating application of AI/analytics to longevity research is the use of these technologies to assess the biological age of human brains. Findings have important implications for the lifestyle and career choices of patients, as well as the diagnosis and treatment of depression, autism, dementia, Parkinson's, and a range of other conditions. Mindstrong Health, a Palo Alto-based company, is developing objective biomarkers of brain function that rely on patterns of interaction between users and their smartphones.  Mindstrong's AI-based app allows the company to assess enormous volumes of user data. Especially interesting is the ongoing effort to determine what is—and is not— “normal” brain aging. Researchers in the UK recently collected brain data from the MRIs of more than 50,000 patients between the ages of three and 96 to help establish, using AI, a “normal brain aging trajectory.” This was only possible because AI proved so adept at harmonizing data gleaned from different types of MRIs performed on different patients at different times and in different locations. By then comparing data from thousands of patients with healthy brains and various brain disorders to the AI-defined “normal” brain aging patterns, the authors were able to draw conclusions about how common brain disorders affect brain aging. Remarkably, AI proved perfectly capable of comprehending, and adjusting for, different rates of aging within individual brains (i.e. between different brain regions). Ultimately, AI allowed the researchers involved to conduct a much bigger and broader study than would have been the case otherwise. The results have profound implications for patient monitoring and treatment, including the tracking of brain and neurological disorders over time.

It is increasing clear that AI is proving indispensable to the process of collecting and analyzing patient data for the purpose of improving human health and increasing human longevity. Most of us recognize AI's role in powering virtual assistants (Alexa, Siri et al.) and the importance of these and future technologies in maximizing quality of life—especially for the elderly. After all, these technologies can help ease the burden on aging brains and bodies and make it more likely that all of us, as we age, live well and live longer while remaining in their own homes. But it is AI's ability to identify and assess various biological markers of age—a much more accurate measure of health status than chronological age—that represents the technology's greatest contribution to increased human longevity. At present, most of the bio-marker data AI uses or generates is useful to entities concerned more with the “big picture” than with individual patient outcomes. Insurance companies, pharmaceutical companies, and research institutions are especially enthusiastic about AI's power to parse bio-marker data for their own purposes. Sooner rather than later, however, AI's chief value will be in delivering the most efficient, personalized, and cost-effective health care possible—across the board—to every patient who needs it. Ultimately, a complete, AI-mediated understanding of biological age—as opposed to chronological age—is integral to maximizing human health and happiness and, ultimately, human longevity.