Plasma proteomic signatures of cellular aging predict human disease.
Level 3 - non-randomized controlled study
Prospective cohort study analyzing plasma proteomic biomarkers to predict incident disease and 15-year mortality
PubMed 42297981 · doi:10.1038/s41591-026-04446-y
What was done
Machine learning models were developed using measurements of over 7,000 plasma proteins from 60,542 individuals to estimate the biological age of over 40 cell types across neuronal, immune, glial, endocrine, epithelial, and musculoskeletal lineages. The models were evaluated for associations with genotype, disease status, and prediction of incident disease and all-cause mortality over 15 years of follow-up.
What was found
Accelerated aging in a single cell type occurred in 20-25% of individuals, and 1-3% showed accelerated aging across 10 or more cell types. APOE4 carriers showed older astrocytes and younger macrophages compared to APOE3 carriers, with inverse associations for APOE2. In individuals with two APOE4 alleles, extreme astrocyte aging tripled the risk of incident Alzheimer's disease. Extreme skeletal myocyte aging was associated with a 12.7-fold higher risk of amyotrophic lateral sclerosis. In smokers, extreme respiratory epithelial aging was associated with a 58% higher lung cancer risk compared to smoking alone. A polycellular aging risk score stratified mortality risk across cohorts.
Why it matters
This work demonstrates that circulating plasma proteomics can non-invasively estimate aging trajectories across specific cell types. These cell-specific biological clocks identify individualized vulnerability to distinct age-related diseases and overall mortality.
Limits
The abstract does not provide confidence intervals, absolute risk numbers, or model performance metrics (e.g., C-statistics, AUC). Cell type-specific aging scores are computational inferences from circulating blood proteins rather than direct histological or cellular measurements. Participant demographics, ancestry, and replication cohorts are not detailed in the abstract.