New Facts About the Genetics of Longevity Uncovered
For decades, statistical noise masked the genetic drivers of human aging. Now, a mathematical correction has revealed that our intrinsic lifespan is highly heritable—and driven by reversible biological software.
The information in this article is for educational purposes only and is not intended as medical advice. Always consult a qualified healthcare professional for medical questions.
Summary
This article reveals how mathematics, epigenetics, and AI are reshaping our understanding of human longevity. Readers will learn that biological aging is not an inevitable decline, but a highly heritable yet modifiable process. Key takeaways: - **True Heritability:** By filtering out accidental deaths, researchers found our "intrinsic" lifespan is 55% heritable—double previous estimates. - **Reversible "Software":** Aging is largely driven by reversible epigenetic glitches rather than permanent DNA damage. - **AI Forecasting:** Advanced AI models can now predict cellular aging trajectories and measure the "brain age gap." - **Actionable Aging:** Both cellular reprogramming and healthy lifestyle choices can actively slow or reverse biological aging.
Two identical twins are born in the early 20th century. They share the same genetic code. At age 25, one is struck by a runaway carriage and dies; the other lives to be 85. In this extreme case, the external event completely overshadowed any genetic influence on their longevity, making it appear that their shared genetics exerted no influence on their lifespans.
Historical studies of human longevity were often complicated by extrinsic mortality—deaths from factors like accidents, homicides, infectious diseases, and environmental hazards—which contributed to a statistical illusion that cast doubt on the role of genetics in aging according to a 2026 study. Based on vast registries of twins and sprawling family pedigrees, historical estimates suggested genetics played a limited role in human lifespan, with twin studies typically finding a 20 to 25 percent heritability, and large pedigree studies reporting even lower figures, sometimes as low as 6 percent. These low estimates contributed to skepticism about the feasibility of identifying genetic determinants of longevity.
However, recent findings are prompting a re-evaluation of previous estimates. By mathematically filtering out "extrinsic" deaths—those caused by accidents, infections, and other outside forces—researchers have revealed that the heritability of our "intrinsic" lifespan is actually around 55 percent. This mathematical correction does more than revise a number; it reframes longevity genetics as a more promising field, aligning human longevity with the heritability of other complex traits.
Concurrently, while the intrinsic heritability of human lifespan is estimated at about 50%, biologists and computer scientists are exploring evidence, particularly from mammalian models, that aging may be significantly shaped by reversible "software" glitches in the epigenome, rather than solely by irreversible "hardware" damage to DNA. Supercharged by artificial intelligence models capable of modeling cellular state trajectories across the human lifespan, these findings suggest that aspects of aging, such as brain aging, may be computationally predictable and potentially modifiable as demonstrated by studies on the brain age gap, rather than solely a deterministic march toward death.
The Statistical Illusion
To understand how longevity genetics lost its way, one must look at how heritability is calculated. Heritability is not a fixed property of a trait; it is a measure of how much of the variation in a trait within a specific population can be attributed to genetic differences. If a population is subjected to high rates of random, environmentally driven mortality, the genetic signal is drowned out by the noise.
Using mathematical analysis and advanced simulations, researchers recently reexamined historical datasets, including Danish, Swedish, and SATSA twin cohorts, alongside U.S. centenarian sibling data. They found that historical cohorts with high extrinsic mortality severely compressed observed twin correlations. When a significant portion of a population dies from infectious diseases or accidents before their biological clocks run out, the data falsely suggests that genetics do not matter.
The researchers demonstrated that when extrinsic mortality is corrected for, the heritability of intrinsic lifespan jumps to about 55 percent. They also identified a nonintuitive, nonlinear effect regarding the minimum age cutoffs used in these studies. In historical populations with high extrinsic mortality, setting a high cutoff age (e.g., only studying individuals who lived past 60) helps filter out the noise of early accidental deaths, revealing the genetic signal. However, in modern populations with low extrinsic mortality, setting the cutoff age too high artificially truncates the natural variance in lifespan, paradoxically lowering the heritability estimates.
By proposing a standardized definition of "intrinsic lifespan heritability," the researchers have provided a new lens for aging biology. If intrinsic lifespan is around 55 percent heritable, it suggests a more substantial genetic influence on longevity than previously thought. Instead, it underscores the strong biological and genetic determinants of intrinsic lifespan, making the study of longevity genes a more promising field for revealing aging mechanisms, the researchers suggest.
Hardware vs. Software
This mathematical revelation naturally begs a biological question: If intrinsic aging is highly heritable, what exactly is being inherited?
A parallel line of research challenges the traditional paradigm that mammalian aging is primarily driven by the accumulation of genetic mutations—irreversible "hardware" damage to the DNA itself. Instead, researchers are finding strong evidence for the Information Theory of Aging, which proposes that aging is driven, at least in part, by the progressive loss of epigenetic information—a "software" problem—though the full extent of its role compared to "hardware" damage is still being investigated.
The epigenome is the complex system of chemical tags and proteins that tells a cell which genes to turn on and off, dictating whether it functions as a skin cell, a neuron, or a muscle cell. As cells constantly repair double-stranded DNA breaks caused by normal metabolic processes, chromatin-modifying proteins are temporarily drawn away from their usual locations to assist in the repair. Over time, this repeated relocalization degrades the cellular Waddington landscape—the metaphorical hills and valleys that keep a cell's identity stable. The cell forgets what it is supposed to be.
To prove this, researchers developed the "ICE" (inducible changes to the epigenome) mouse model. Using a highly specific endonuclease called I-PpoI, they induced non-mutagenic DNA breaks in the mice. This forced the cells to undergo faithful DNA repair without altering the underlying genetic code. The result was a rapid acceleration of the epigenetic clock. The mice exhibited classic hallmarks of aging across multiple tissues, including cognitive decline, muscle weakness, frailty, and histological degradation in the kidneys and skin, all without accumulating new DNA mutations.
Crucially, because this degradation is a software problem, it can be rebooted. The study demonstrated that these age-related changes could be reversed using OSK-mediated epigenetic reprogramming. By expressing a specific set of transcription factors, the researchers reset the epigenome, restoring cellular identity and function. Mammalian aging, it appears, can be driven both forward and backward.
The Cellular Time Machine
If aging in mammals can be understood as a reprogrammable software process, driven by epigenetic changes as suggested by recent research, a key challenge is tracking and predicting that software degradation across a human lifespan. Many previous models in network biology have struggled to learn how cellular responses unfold over time, particularly across the long trajectories relevant to human aging, because they typically considered only one cell state at a time according to one study.
To bridge this gap, computational biologists are turning to deep learning. Researchers recently developed MaxToki, a foundational temporal AI model trained on nearly 1 trillion gene tokens. Unlike previous models in network biology that analyze single, static snapshots of cells, MaxToki is designed to forecast long-term cellular trajectories.
The model was trained on a massive population-based corpus of approximately 22 million cells spanning human life from birth to over 90 years old. Its architecture relies on a sophisticated two-stage training process. First, it learns to generate single-cell transcriptomes using a rank-value encoding that prioritizes highly dynamic genes. Second, its context window is expanded using RoPE (Rotary Position Embedding) scaling, allowing it to process multiple cells across a timeline.
Using continuous numerical tokenization, MaxToki successfully learned to predict the time elapsed between cell states. It can generate accurate future, past, or intervening cellular profiles, even for ages and cell types it has never seen before. During rigorous validation, the model generated biologically faithful, single-cell resolution states that fooled external classifiers, and it successfully captured the complex, non-monotonic gene changes that occur during cellular reprogramming. MaxToki effectively serves as a cellular time machine, enabling large-scale in silico screening to discover interventions that could slow or reverse age-related diseases.
Minding the Gap
While AI models like MaxToki map the microscopic trajectories of single cells, other deep learning systems are tracking the macroscopic consequences of aging in human organs.
A recent study investigated the clinical utility of the "brain age gap" (BAG)—the difference between a person's biological brain age and their chronological age. To measure this, researchers deployed a novel 3D Vision Transformer (3D-ViT) deep learning model trained on MRI scans from over 40,000 participants across three major cohorts, including the UK Biobank. The model achieved brain age estimations with a mean error of 2.68 years in the UK Biobank and 2.99–3.20 years in the ADNI/PPMI cohorts.
The findings position the brain age gap as a powerful, dynamic biomarker for neuropsychiatric health and mortality. The researchers found that each one-year increase in BAG is associated with a 16.5 percent higher risk of Alzheimer's disease and a 12 percent increase in all-cause mortality. Individuals in the highest risk quartile face dramatically elevated risks, including a 2.8-fold increase for Alzheimer's and a 6.4-fold increase for multiple sclerosis, alongside notable declines in cognitive processing speed and reaction time.
Yet, much like the epigenetic software of the ICE mice, the brain age gap is not a fixed destiny. The study demonstrated that brain aging trajectories are highly modifiable. Healthy lifestyle interventions—specifically smoking cessation, moderate alcohol consumption, and regular physical activity—were shown to significantly slow BAG progression, even in individuals with advanced neurodegeneration.
Advances in fields like mathematics, epigenetics, and artificial intelligence are collectively deepening our understanding of biological aging. By filtering out the noise of extrinsic mortality, researchers have refined estimates of the genetic contribution to intrinsic lifespan, suggesting it is about 55% heritable when confounding factors are addressed. While this substantial heritable component is being clarified, the field is also exploring how biological aging may involve dynamic, reprogrammable systems, particularly through epigenetic mechanisms as demonstrated in mouse models. OSK-mediated cellular rejuvenation has shown promise in mouse models by reversing age-related changes, while studies indicate that specific lifestyle modifications, such as smoking cessation, moderate alcohol consumption, and physical activity, can significantly slow the progression of brain aging in humans by slowing the progression of the brain age gap, particularly in individuals with advanced neurodegeneration. Furthermore, an advanced AI model like MaxToki is beginning to predict cellular trajectories and identify potential interventions to modulate age-related changes. Together, these emerging insights suggest that while intrinsic lifespan has a significant heritable component, certain aspects of biological aging may exhibit a potential for modifiability—not merely an inevitable decline, but a biological process with aspects that could potentially be influenced or, in specific contexts like cellular reprogramming, even reversed or rewritten, particularly in areas like brain health.
Also in Aging
The Wrong Suspect in the Cell’s Burn Chamber
For decades, scientists blamed the Krebs cycle and fat-burning machinery for making mitochondria leak toxic peroxide. But under moderate beta-oxidation, the usual suspects are exonerated. The real mystery is not why the cell is burning fuel—it’s what else inside the mitochondrion is making the dangerous exhaust.
