Plasma proteomic signature of risk and prognosis of frailty in the UK Biobank.

Abstract

The availability of proteomics data in large, population-based cohort studies offers an unprecedented opportunity to understand the onset and progression of aging-related diseases and syndromes. We examined the proteomic signature of the onset of frailty and the progression to death among the prefrail and frail. A total of 2920 proteins were assayed using Olink among 43,895 participants (aged 39-70 years) in the UK Biobank. Using multinomial logistic models, we identified 102 and 90 proteins cross-sectionally associated with baseline prefrailty and frailty (Bonferroni-corrected p-value < 0.05), respectively. Additionally, cox regression identified 87 and 48 proteins associated with death among initially prefrail (n = 16,661) and frail (n = 1647) individuals, respectively. Eight overlapping proteins were cross-sectionally associated with prefrailty and frailty at baseline and prospectively associated with death among prefrail and frail individuals. CD300E, GDF15, and PLAUR were the most significant proteins among these eight proteins. LASSO regression selected 73 and 23 proteins predicting death in prefrail and frail participants, respectively. Protein-based prediction models based on LASSO regression and the light gradient boosting machine classifier demonstrated satisfactory discrimination, calibration, and reclassification among the prefrail and frail. GDF15, WFDC2, and NEFL were the most important proteins predicting death among prefrail and frail individuals. Proteins associated with the onset and progression of prefrailty and frailty were enriched in pathways involving protein metabolism, cellular signaling in disease, apoptosis, and inflammation. Our research could uncover novel therapeutic targets for addressing the onset and progression of frailty, potentially informing the design of patient-centered treatment strategies and management plans.