Navigating aging: from animal models to AI breakthroughs

At the turn of the century, both Craig Venter and Steve Jobs agreed on the 21st century being “The Century of Biology." By now, this realization has become commonplace way beyond biologists, the medical establishment, and pharma, as we witness tech entrepreneurs, capital allocators, and policymakers all gearing up to claim their chunk of the innovation in the fields of biology and biotechnology and their enormous anticipated applications to improve human health and existence.

Of all the applications of biotech and life sciences, one stands out as a final frontier that can upend and upgrade human existence like nothing before it. And that is the potential to defy, delay, and potentially reverse the aging process itself. In this case, longevity is the biological equivalent of nuclear fusion, unlocking the most finite of all resources, time itself.

To get more practical, one of the starkest differences between aging and almost any other biology problem has to do with the concept of biological mechanism, or the lack thereof. The conventional study of biology focuses on understanding biological mechanisms for basic biology and the malfunctioning thereof (pathophysiology) to establish a better understanding of diseases and potential cures.

In contrast, aging is a phenomenon that occurs invariably with the passing of time, albeit at varying speeds in different individuals. Classical medicine is yet to recognize aging as a disease and regulate the body to recognize it as a “druggable” indication in humans (a startup has secured a special indication for the use of their novel molecule to slow down aging in large dogs).

Semantic and technical jargon aside, the core difference between aging and any other individual disease is the lack of a “definable” mechanism of aging. This is a hard science problem, which I would bet would prove as essentially difficult as the uncertainties of quantum mechanics, and these two crucial points justify this conviction.

Ageing did not “evolve” as a function for its own end; in other words, nature never selected for aging; it simply never selected against it, as natural selection is only conserved with passing the germline and hence only optimized our existence up to the point of procreation and ensuring the survival of our progeny.
The lack of an “evolutionary” route of aging probably means it is a stochastic entropic process, the culmination of millions of things that can go wrong; in other words, the integer of all minute imperfections in the otherwise spectacular machinery of biology.

With that in mind, the field of aging and longevity has identified an expanding list of “hallmarks of aging.” Each is correlated with the passing of time, but none is exclusive as the “root cause” of biological aging, and causal relationships between these remain elusive and complex. Built on the hallmarks are “biomarkers” or aging, also called “aging clocks," which can tendentially identify the speed and acceleration of biological aging in various individuals but are in no way reliable enough as an absolute quantitative measure, especially for comparing different individuals or cohorts.

A very practical implication of the lack of a clear mechanism and direct targets for the aging process is the difficulty of generating animal models faithfully representing the aging process to test interventions for their potential to delay or reverse the aging process. Coupled with the lack of reliable biomarkers, this makes for two-fold problems: limiting the reliability of most interventions that have been studied in inadequate models and having endpoints (biomarkers) that do not necessarily predict lifespan.

One of the biggest challenges comes from the status- quo of biomedical research relying on homogenous inbred rodent strains; these have become dominant in research due to factors irrelevant and contrary to the scientific reliability and relevance to humans or any other “naturally occurring” animals. The most popular rodent strains dominated the industry due to economic factors, including fast breeding, rapid growth, and a shorter lifecycle. Science has overrelied on these abundantly available animal models due to convenience, and homozygous animal populations (that are inbred animals) have become favored as a homogenous genetic background ostensibly makes results easier to analyze and replicate. The major problem here is that results from a particular inbred animal strain are rarely seen in other inbred strains and are ever less likely to apply to wild-type mice, let alone humans or primates in general.

These are all serious shortcomings that the fields of preclinical and academic biomedical research, including cancer research, neurodegenerative diseases, and gene delivery, are all struggling with. Nonetheless, in the field of aging and longevity, these shortcomings render such inbred animal models completely useless, as the homozygous genetic background always correlates to predisposition to particular diseases based on their particular mutations, which in turn makes it impossible to entangle whether an intervention extends lifespan by delaying aging itself or simply preventing a particular disease to which the inbred animal population is predisposed. In fact, cancer is almost the exclusive cause of death.

Therefore, in the absence of a singular target, pathway, or mechanism of aging, the challenge is to establish animal models that can be reliably used to find and validate interventions that can delay aging and prolong a healthy lifespan. There are two broad attributes that help recapitulate the aging process using research models to allow the identification of interventions with real anti-aging potential.

Research models that mirror the natural aging process within a context most relevant to humans. Factors to optimize for include:
- Heterozygosity and genetic heterogeneity of the animal population.
- Healthy living conditions of the animals, including physical activity and a balanced diet, as opposed to ad libitum feeding within tight cages.
- The use of “naturally” aged animals, as opposed to “fast” aging models, as in the absence of a unified mechanism of aging, any accelerated aging mode (e.g., by artificially introducing DNA breaks) is probably another disease model with poor prediction power for aging in healthy individuals.
Due to the absence of reliable aging biomarkers, studies should thrive to include a “hard endpoint” as it pertains to lifespan, in other words, to follow up the subjects up until the natural death of all, or a pre-defined part of each cohort. Keeping the aforementioned considerations in mind, there are currently three beacons of hope for achieving rigorous models to test anti-aging and longevity interventions.

Maximizing the reliability of mice in aging research, the intervention testing program ITP: the ITP is a program of the National Institutes of Aging. The program tests anti-aging candidate interventions on the UM-HET3, a heterozygous mouse strain with reproducible genetic heterogeneity, simultaneously in three independent laboratories. Furthermore, it includes testing in mid-aged mice and follows up to reach the hard lifespan outcome with rigorous predefined statistical rules and cut-off criteria to identify successful interventions. With all its rigor, the ITR still leaves much to be desired as pertains to the lifestyle, and functional assessment of the mouse cohorts. However, these perceived shortcomings are understandable in light of the larger numbers of mice tested to allow enough statistical power. In my opinion, the ITP sets a remarkable example of methodical scientific rigor that should be extended beyond the aging and longevity field to other areas of biomedical research plagued by unreliable and unpredictive homozygous animal models. The longevity community should build upon this initial success example by stressing the importance of rigorous animal testing to identify interventions that can extend the lifespan of a general population of healthy people, as opposed to prolonging lifespan only in people with specific disease predispositions. This activity is not restricted to academia, as longevity funders, CROs, and founders all have a role in promoting and ensuring the availability of improved animal models and testing protocols.
Companion animals as a target for anti-aging interventions: this remarkably bridges the gap between preclinical and clinical research, as extending the healthy lifespan of companion animals is a worthwhile pursuit and a great go-to-market opportunity for longevity interventions. The work on companion animals builds on preclinical research done in lab animal models, and by doing so, it is doing a great service to the field of human longevity by substantiating the validity of anti-aging interventions in companion animals that are much closer to humans than mice in their genetics, size, average lifespan, environment, and lifestyle. The Dog Aging Project, led by Matt Kaeberlein, has delivered one of the earliest successes in this field by showcasing the preliminary potential of rapamycin to improve health span in dogs and continues to seek the actual lifespan endpoint of rapamycin treatment in dogs in the ongoing TRIAD study. More remarkably, the company Loyal has done a remarkable service for the longevity field, getting the first ever official recognition from the FDA of lifespan extension as an indication for drug discovery and development. This came in the form of the FDA Center for Veterinary Medicine granting an expanded conditional approval of Loyal’s LOY-001 for lifespan extension in larger dogs. Personally, I am very passionate about this sector of the longevity field, and I think it should be extended to other companion animals that can both immensely benefit from and add to our insights into how to live longer and healthier.
This is not exactly an "animal,” but it still cannot be ignored in our day and age. AI models are of immense value, both for untangling the complexity of aging and as in-silico models to replace animals. When reckoning with aging as a complex, hyperdimensional phenomenon, AI cannot be dispensed with as a potential key to both solving the hard aging problem “assuming it is solvable” and the more practical and immediate use of building predictive models to allow the prediction of successful anti-aging interventions. A remarkable example of the first use of AI to develop a “theory of aging” on the way to solve the aging hard problem is the work done by GERO led by Peter Fedichev, who recently shared a preprint study entangling the dynamic and entropic aspects of aging based on single-cell DNA methylation patterns in cells. And a more practical breakthrough in the use of AI to predict anti-aging targets and interventions recently came from Insilico Medicine, led by Alex Zhavoronkov, as their AI-identified novel antifibrotic target TNIK has also been indicated as highly relevant to multiple hallmarks of aging. As in every other field of human activity, AI is not to be glossed over when considering its potential, and it might even turn out to be the singular solution to the very complex, unstructured, and amorphous aging problem.

Conclusion

Identifying better aging biomarkers and establishing rigorous models for testing anti-aging interventions and for untangling the aging phenomenon go hand in hand to accelerate progress in the longevity and healthspan fields and are both musts for translating the potential of preclinical research into actionable interventions that can make a difference for humanity in our lifetime.