Sleep chart of biological aging clocks across organs and omics

Adequate, regular sleep supports healthy aging and longevity. This work introduces a “Sleep Chart” mapping self-reported sleep duration to 23 biological aging clocks spanning 17 organ systems or tissues across three omics layers (imaging, plasma proteomics, plasma metabolomics). Across nine brain and somatic systems and all three omics, both short sleep (<6 h) and long sleep (>8 h) associate with higher biological age gaps (BAGs) in a systemic U-shaped pattern; optimal sleep differs by organ and sex (roughly 6.4–7.8 h). Relative to normal sleep ([6–8] h), short and long sleep also align with higher risks of extracerebral diseases and all-cause mortality, supported by genetic correlation and time-to-event analyses (e.g., migraine, depression, diabetes). Short vs long sleep relate to late-life depression through different pathways: long sleep is more plausibly mediated by biological aging, whereas short sleep links more directly to outcomes. Mendelian randomization does not support pervasive causal effects of disease on sleep disturbance, yet sleep may still partly reflect latent disease burden. Overall, the U-shape is more consistent with modifiable sleep issues than pure genetic susceptibility, suggesting sleep optimization may support healthier aging, lower disease risk, and longer life. Interactive portal: https://labs-laboratory.com/sleepchart.

Multi-organ, multi-omic biological age gaps (MRIBAG, ProtBAG, MetBAG, etc.) capture aging beyond calendar age. Prior work hints at U-shaped sleep–clock relationships; here, in UK Biobank, we test whether the pattern generalizes across organs, omics layers, and sex, integrating FinnGen, PGC, TriNetX, MESA, and more under the MULTI framework to connect sleep disturbance with genetics, incidence, mortality, and late-life depression pathways.

• Sleep vs BAG: self-reported duration (field 1160), window 4–10 h; GAMs (mgcv, cubic splines) for 23 BAGs, 720 IDPs, 2,923 plasma proteins (342 organ-enriched), 327 metabolites (107 organ-related); covariates include demographics, anthropometrics, blood pressure, assessment center, disease indicators, and (some models) 40 genetic PCs; Bonferroni within analysis families.
• Genetics: European-ancestry UKBB; REGENIE GWAS for short vs [6–8] h and long vs [6–8] h; FUMA loci, MAGMA (GTEx v8), LDSC genetic correlation with 527 FinnGen/PGC traits; extended data on BAG polygenic scores vs sleep.
• Survival: Cox models for 726 de-novo ICD-10 endpoints and all-cause mortality; TriNetX validation using insomnia/hypersomnia as proxies for short/long sleep.
• Mediation: structural equation models (lavaan, bootstrap) for sleep → MRIBAG → LLD1/LLD2; reverse specifications and disease→sleep MR sensitivity analyses.

Nine of 23 BAGs show significant nonlinear (U-shaped) associations with sleep (P < 0.05/23), spanning brain, lung, liver, immune, and skin ProtBAGs, endocrine MetBAGs, and brain, adipose, and pancreatic MRIBAGs; optimal sleep is roughly 6.4–7.8 h (women) and 6.4–7.7 h (men). The pattern extends across 720 IDPs, 342 organ-enriched proteins, and 107 organ-related metabolites. Downstream analyses bin sleep as short (<6 h), normal [6–8] h, long (>8 h), with sensitivity analyses using [7–9] h as “normal.”

GWAS for abnormal vs normal sleep yields five genome-wide significant loci; MAGMA shows short-sleep genetics enriched in multiple brain regions. BAG PRS–sleep relationships largely lose the U-shape, suggesting environmental drivers with partial genetic overlap for some endocrine/metabolic clocks. LDSC links abnormal sleep to 153 disease traits (short sleep more systemic; long sleep more neuropsychiatric). Cox models surface 153 significant disease associations (short sleep dominant; HRs ~1.2–7.0); all-cause mortality HRs vs [6–8] h are ~1.50 (short) and ~1.40 (long). TriNetX aligns with UK Biobank conclusions.

Structural equations from baseline sleep → follow-up MRIBAG → LLD subtypes suggest short sleep acts more through direct paths toward LLD, with adipose MRIBAG among mediators; long sleep leans more on organ mediators (especially brain MRIBAG). Reverse MR does not support broad causal disease→sleep effects; bidirectional confounding cannot be fully excluded.

The Sleep Chart situates sleep duration alongside multi-organ, multi-omic aging burden and systemic outcomes as a plausibly modifiable exposure. Limitations include self-reported sleep, predominantly European GWAS, residual confounding (long sleep may flag subclinical illness), and the need for wearables and independent cohorts. Full methods, figures, and interactives are on the Sleep Chart portal and in the PMC full text.

• Sleep Chart portal — summary statistics, plots, interactive browsing
• Genetic correlation browser sleep_gc.html · Cox visualization sleep_cox.html
• MEDICINE — GWAS for 23 BAGs
• Synapse: syn64923248, syn68602065
• Software stack per paper and PMC supplements (MLNI, mgcv, GAMLSS, lifelines, lavaan, STRING, MetaboAnalyst, REGENIE, LDSC, FUMA, etc.).

After Nature publication, cite the journal version. Until then, medRxiv + PMC:

The MULTI study et al. Sleep chart of biological aging clocks across organs and omics. medRxiv (2025). 10.1101/2025.08.08.25333313; PMID 40832429; full text PMC12363694. Accepted by Nature (in press).