ProtPhenoAge: a plasma-proteomics aging clock predicts mortality in 37,433 UK Biobank adults
ProtPhenoAge, an aging clock built from 185 plasma proteins in 37,433 UK Biobank participants (Wang et al., J Adv Res, 11 May 2026; DOI 10.1016/j.jare.2026.05.022, PMID 42119967), predicted all-cause mortality (AUC 0.76) and outperformed chronological age, PhenoAge and ProtAge; its acceleration linked to 313 disease phenotypes, with Mendelian-randomization and GWAS support (APOE rs429358). A net-new proteomic biomarker — strong for research, but internally validated only, not a clinical endpoint.
Aging clocks are the measurement layer the whole longevity field depends on. ProtPhenoAge is a new one built from the proteome rather than DNA methylation. According to the study (Wang et al., Journal of Advanced Research, 11 May 2026; DOI 10.1016/j.jare.2026.05.022, PMID 42119967), the authors used Boruta-SHAP feature selection to pick 185 PhenoAge-related plasma proteins from Olink Explore data in 37,433 UK Biobank participants (median follow-up ~13.7 years), then trained an XGBoost model. ProtPhenoAge correlated tightly with PhenoAge (r = 0.96) and predicted all-cause mortality at AUC 0.76 — beating chronological age, PhenoAge and the earlier ProtAge clock. Its acceleration associated with 313 disease phenotypes across a phenome-wide scan, and GWAS/colocalization flagged aging loci including APOE (rs429358), giving it a genomic anchor most clocks lack.
Why it matters (money + metacrisis lever): a credible, mortality-predictive, blood-based aging biomarker is the substrate for healthspan trials (a readout that moves in years, not decades) and for a fast-growing consumer "biological age" market. It is simultaneously the place where measurement hype most outruns evidence — precisely the seam RDKG/MKG exists to police. Proteomic clocks are also more mechanistically legible (druggable proteins) than methylation clocks.
Conflict-of-interest note: an academic study (Nanjing Medical University), funded through public/academic channels with no disclosed commercial product — low direct COI. The relevant risk is downstream: consumer-testing firms appropriating "aging clock" language to sell unvalidated "biological age" scores as clinical results.
What would prove it wrong: ProtPhenoAge fails to replicate in an external, non-UK-Biobank, multi-ancestry cohort, or does not predict morbidity/mortality better than existing clocks after adjustment — or fails to move in response to interventions that demonstrably extend healthspan.
Live-status check (as of 2026-07-07): published, peer-reviewed, in-window (11 May 2026). Evidence is a single large cohort with internal train/validation; external prospective validation and intervention- responsiveness are not yet established. Not a diagnostic; not a surrogate endpoint accepted by any regulator.
Informational only — not medical advice. Cite the primary paper (DOI above); do not reproduce.
Log ProtPhenoAge as a verified, in-window addition to the aging-biomarker cabinet alongside DunedinPACE — usable as an exploratory trial readout — while refusing to sell any single "biological age" score as a clinical endpoint until external, prospective validation lands. The validated-vs-hyped biomarker line is the product.
domains: aging-clock · biomarkers · proteomics · mortality · uk-biobank · longevity