Network-based Plasma Proteomics Reveals Molecular Overlap Between Physical Activity and Dementia Risk

Physical activity (PA) is linked to lower dementia risk, but molecular pathways underpinning PA-related dementia risk are poorly understood. We conducted plasma proteomics (SomaScan v4.1) and 30-day Fitbit-based PA monitoring (average daily step count) in 65 cognitively unimpaired older adults from the UCSF BrANCH cohort. Differential regression and network analyses identified PA plasma proteomic signatures tied to extracellular matrix (ECM), immune response, and lipid metabolism. Protein module M12 ECM/neurodevelopment positively correlated with PA in BrANCH and external cohorts, inversely predicted cognitive aging outcomes in BrANCH, and decreased across multiple neurodegenerative conditions. M12 was enriched for proteins from Alzheimer’s disease (AD) risk genes and antemortem plasma abundance of ANTXR2, an M12 ‘hub’ protein, forecasted longitudinal cognitive decline and postmortem brain tissue protein signatures of AD cognitive resilience in the ROSMAP cohort. Our integrated analysis across six proteomic datasets identified blood-detectable molecular signatures of PA and neurodegenerative disease, including ECM-related proteins (e.g., ANTXR2) that may represent key molecular targets for dementia prevention.