Abstract
Cerebrospinal fluid (CSF) biomarkers are central to Alzheimer's disease (AD) diagnosis and research. However, CSF composition is shaped not only by neurodegeneration, but also by underlying physiological and pathological processes that remain poorly characterized. By integrating multi-omics data from the deeply characterized memory-clinic ACE CSF cohort (N=1,372), the Global Neurodegeneration Proteomics Consortium (N=1,863), and publicly available quantitative trait loci data, we reveal that 73.2-85.9% of the molecular variance in CSF omics data is driven by two main factors: one reflecting CSF turnover rate, and another representing blood-brain barrier (BBB) integrity. CSF turnover mainly determines brain-derived molecules, while BBB damage leads to increased blood-derived protein abundance. CSF turnover/clearance severely impacted core AD biomarker levels, affecting the classification of subjects in the A/T framework. Adjusting biomarker levels for OPCML, a novel reference marker, improved biomarker-based prediction of AD progression and removed confounded associations, revealing a proteomic signature of sporadic AD pathology that closely resembles that of autosomal dominant AD. Finally, using the ACE CSF cohort as discovery (N=1,221) and Knight ADRC as replication (N=1,073), we report a curated AD signature comprising 446 unique proteins. Our findings identify CSF dynamics as a major source of molecular variation, reshaping the interpretation of CSF biomarkers.