Hypothesis

If aging arises from declining selection pressure rather than an adaptive death program, then genes that sit at the intersection of high early‑life essentiality and relaxed late‑life constraint will form topological hubs in the human interactome. These hubs should show (i) strong purifying selection on their early‑life interacting partners, (ii) weaker conservation of late‑life‑specific interactions, and (iii) a measurable fitness trade‑off: perturbing the late‑life‑specific edges extends lifespan without reducing early‑life reproductive output.

Mechanistic reasoning

1. Evolutionary expectation – Under mutation accumulation and antagonistic pleiotropy, selection removes deleterious alleles that affect fitness before reproduction, but allows late‑acting deleterious alleles to persist. Genes whose products are required for growth, development, or early‑life reproduction become highly connected “date hubs” that coordinate multiple pathways. Late‑life‑specific interactions, being under relaxed constraint, accumulate more sequence divergence and are more likely to be condition‑dependent (e.g., stress‑responsive).
2. Network prediction – A graph‑neural‑network (GNN) trained on experimentally validated aging interventions from the NIH/NIA Interventions Testing Program NIH/NIA ITP should learn to distinguish two classes of nodes:
   • Programmed‑like: densely connected modules with uniform high conservation across all ages.
   • Non‑programmed: hubs with high early‑life partner conservation but a distinct periphery of low‑conservation, age‑specific links.
3. Experimental test – For top‑scoring non‑programmed hub genes (e.g., HSP90AA1, TP53, AKT1), use CRISPR‑based allele‑specific editing or inducible RNAi to selectively disrupt the late‑life‑specific interactions (identified by age‑stratified co‑expression or phosphoproteomics) while preserving the early‑life interaction network. Measure:
   • Early‑life fitness: fecundity, developmental timing, and stress resistance in young adult C. elegans, Drosophila, or human iPSC‑derived organoids.
   • Late‑life phenotypes: lifespan, healthspan markers (motility, proteostasis, epigenetic clock). Prediction: Lifespan extension ≥15 % with no significant decline (p>0.05) in early‑life fitness relative to controls.

Falsifiability

• Null outcome: If disrupting late‑life‑specific edges fails to extend lifespan or invariably reduces early‑life fitness, the hypothesis that aging is a byproduct of relaxed selection on peripheral interactions is weakened, suggesting either a more global pleiotropic architecture or an adaptive death program.
• Alternative outcome: If lifespan extension occurs only when the entire hub is knocked down (affecting both early and late interactions), this would support the idea that aging stems from core pleiotropic genes rather than separable late‑life modules, still compatible with non‑programmed models but indicating a different topological signature.

Broader implications

Confirming this hypothesis would shift target selection in longevity medicine from blunt inhibition of conserved “death” proteins to precision editing of context‑dependent interaction interfaces. It would also provide a computational pipeline—GNN classification followed by age‑specific interaction mapping—to prioritize interventions that work with, rather than against, the evolutionary logic shaping our genomes.

References

• ’s critique of programmed aging theory: PMC6398523, PMC8813929
• NIH/NIA Interventions Testing Program curated targets: programmed-aging.org