Proteomic inflammaging score outperforms conventional markers in predicting all-cause mortality
Investigators developed a proteomic inflammaging score (PIS) using machine learning on a UK Biobank cohort of 40,471 participants and validated it in an external group of 5,250. The final 93-protein signature was positively associated with all-cause mortality and age-related diseases, achieving a C-index of 0.758 when combined with clinical indices and significantly outperforming standard inflammatory markers. The study identifies nonlinear proteomic shifts peaking around ages 50, 62, and 67, and links the score to specific genetic variants. For clinical laboratories, this proteomic framework highlights a compact protein panel that may eventually replace or augment standard inflammatory panels in multiplex assays for aging and chronic disease risk stratification.
The original study
Proteomic signatures of systemic inflammation in aging, multimorbidity, and mortality.
- Authors
- Jiang Z, Ling D, Chen Z, Zhou H, Cui Z, Ma X, et al.
- Journal
- Experimental gerontology
- Type
- Journal Article
- PMID
- 42731645
Original abstract
BACKGROUND: Chronic low-grade inflammation is central to biological aging, but routine inflammatory biomarkers capture limited molecular heterogeneity. We aimed to develop a proteomic inflammaging score (PIS) and evaluate its practical utility. METHODS: In 40,471 England participants from UK biobank, we used LASSO to identify proteins associated with six inflammatory biomarkers (CRP, SII, SIRI, MLR, NLR, PLR), and validated them in 5250 non-England cohort. Deep neural networks generated biomarker-specific scores, integrated into PIS via elastic-net Cox regression. We evaluated associations with mortality, age-related diseases, and aging biomarkers, and compared predictive utility (C-index, NRI, IDI) against conventional markers. DE-SWAN analysis was used to characterize nonlinear age-related proteomic changes, and genetics analyses were performed to investigate genetic architecture. RESULTS: A total of 1112 proteins were identified to associated with all six inflammation biomarkers and 93 proteins retained in simplified versions. Both full (HR = 1.88, 95%CI: 1.79-1.98) and simplified (HR = 1.88, 95%CI: 1.79-1.97) PIS were positively associated with all-cause mortality and multiple aging-related phenotypes. PISs outperformed conventional inflammatory and aging biomarkers, significantly improving mortality prediction: for all-cause mortality, clinical indices plus simplified PIS achieved a C-index of 0.758 (95% CI, 0.752-0.765). External validation in the non-England cohort showed comparable predictive performance. Proteomic inflammation crests were near ages 50, 62-63 and 67 years. 25 lead SNPs were associated with PIS, linking PIS to inflammatory traits and aging biomarkers. CONCLUSION: PIS provides a compact, interpretable proteomic measure of inflammaging and captures mortality, multisystem disease burden, and aging-related biology in population-scale cohorts.