Prospective Investigation Unravels Plasma Proteomic Links to Dementia.

Investigating plasma proteomic signatures of dementia offers insights into its pathology, aids biomarker discovery, supports disease monitoring, and informs drug development. Here, we analyzed data from 48,367 UK Biobank participants with proteomic profiling. Using Cox and generalized linear models, we examined the longitudinal associations between proteomic signatures and dementia-related phenotypes. Mendelian randomization analysis was employed to identify causal associations, and machine learning algorithms were applied to develop protein-based models for dementia prediction. We identified 74 proteins significantly associated with the risk of various types of dementia and cognitive functions after Bonferroni correction. Among these, strong associations were observed for growth/differentiation factor 15 (GDF15), glial fibrillary acidic protein (GFAP), and neurofilament light polypeptide (NEFL), across all types of dementia. Additionally, 15 proteins demonstrated significant associations with neuroimaging-defined dementia endophenotypes. Two-sample Mendelian randomization analyses further substantiated causal relationships between dementia-associated proteins and Alzheimer's disease, particularly involving GDF15, proto-oncogene tyrosine-protein kinase receptor Ret (RET), and GFAP. Moreover, we identified three protein modules associated with dementia, primarily linked to immune system processes, angiogenesis, and energy metabolism, providing insights into potential biological pathways underlying the disease. Furthermore, we proposed a ten-protein panel capable of forecasting dementia over a median follow-up period of 8.6 years, achieving an area under the curve (AUC) of 0.857 (95% confidence interval (CI), 0.837-0.876). Our results revealed dementia-associated plasma proteomic signatures, and their causal relationships, notably GDF15-RET signaling with Alzheimer's disease, and proposed a promising protein panel for high-risk dementia screening.