357 ‒ A new era of longevity science: models of aging, rapamycin trials, biological clocks, & more

KEYPOINTS

• 01

  The 'hallmarks of aging' are readouts of one connected network, not 12 separate targets.

  Kennedy, former CEO of the Buck Institute and now running about 35 PIs in Singapore, wrote the 'pillars of aging' companion review and drew lines connecting all the pillars. Interventions like rapamycin improve every hallmark at once. Healthy aging means a homeostatic network that stays responsive. Fixing hallmarks one by one won't work.

• 02

  Inflammation keeps emerging as central—and baseline mTOR creeps up with age.

  Almost every lifespan-extending intervention reduces chronic inflammation, and inflammation dominates the main components of his lab's aging clocks. Baseline mTOR rises in aging stem cells, losing the dynamic range to switch on for a wound or meal and off otherwise. Elevated mTOR feeds chronic inflammation, which feeds mTOR. Good interventions restore youthful range rather than push anything supra-physiologic.

• 03

  Aging looks like linear damage plus an oscillating 'how well you're coping' component.

  Breaking UK Biobank-type data into principal components, one aging component rises linearly—'damage' or accumulated stochastic change. Another oscillates around it with illness, behaviour and drugs. Nearly all current interventions, including everything in Attia's book, act on the oscillation. Kennedy says that could mean 5–10 extra healthy years—'a revolution'—but probably not 150-year lifespans.

• 04

  The 'valley' resilience model explains exponential mortality from linear damage.

  Following Peter Fedichev's work at Gero, health is a ball in a valley. When young the walls are steep, so you're pulled back after junk food or illness. Damage lowers the hills, so the chance of tipping into a failure state rises exponentially, consistent with Gompertz mortality. With about 50 possible failure states, treating one disease barely helps; you must keep all the hills high.

• 05

  Bending the slope may need new kinds of interventions, such as lowering system 'noise' or reprogramming.

  Kennedy wonders whether we should target noise—such as transcriptional noise—rather than how long worms live, since worms are already in a failure state. He expects fitness trade-offs, which timing might avoid. Long-lived yeast mutants mostly wouldn't survive in the wild. Partial reprogramming could replace damaged cells but is early. He's returning to basic discovery science for this.

• 06

  Methylation clocks are overhyped as causal; many data types make equally good clocks.

  Proteomic, microbiome and facial-structure clocks (Jackie Han's facial clock rivals methylation) also predict age; association isn't causation. In NHANES with about 200 months of follow-up, some first-generation methylation clocks predicted mortality worse than chronological age. GrimAge and PhenoAge beat it. Matt Kaeberlein's eight simultaneous commercial tests disagreed even with identical repeats.

• 07

  A ~50-marker routine-lab clock beat every other NHANES predictor of death, and it's actionable.

  About 30 of the markers come from a complete blood count; the rest are standard labs, inflammatory markers and a few self-reports, costing about $300 in Singapore. It outperformed ASCVD risk scores. Principal components flag fixable drivers, such as four in-range markers adding about four years of biological age, roughly 50% higher mortality risk. At 65 in NHANES, about 20% of people needed but lacked treatment, and well-controlled people on medication aged slower than those who needed none.

• 08

  Rapamycin remains the gold-standard small-molecule geroprotector, but probably with modest human effect.

  Weekly dosing that lets trough levels fall likely avoids immune suppression and may even protect against respiratory infection, as Mannick's trials suggested. Kennedy expects it to extend healthy years rather than change the damage slope. His Singapore trial gives about 5 mg weekly for six months to 150–200 healthy 40–60-year-olds, measuring clocks, cytokines, pulse-wave velocity, DEXA, strength and cognition. Attia doubts most of those endpoints will move.

• 09

  Combining many supplements is more likely to cancel out than add up.

  In mouse combination studies Kennedy sees interference more often than synergy—'mixing 20 colours of paint' gives grey. He adds one compound at a time and measures before and after. With $100 million a year instead of about $4 million, he'd run factorial combination trials and map each compound's binding target.

• 10

  Cardiovascular and metabolic disease are the most preventable of the big four; cancer is the most inevitable.

  Cancer requires accumulated mutations plus immune decline, so it may respond least to longevity interventions. Dementia is likely modifiable, but the data are thin. Kennedy criticises Alzheimer's research for spending 30 years on plaque and imagines an epitaph: 'cleared plaques, died of Alzheimer's at 94.'