Biomarker Discovery in Neurodegenerative Disorders

Abstract

Neurodegenerative disorders -- Alzheimer's disease (AD), Parkinson's disease (PD), amyotrophic lateral sclerosis (ALS), and frontotemporal dementia (FTD) -- collectively affect over 55 million people worldwide, yet diagnosis typically occurs years after irreversible neuronal loss has begun, and no disease-modifying therapy has demonstrated unequivocal clinical benefit partly because trials enrol patients too late in the disease course. Blood-based biomarkers that can detect neurodegeneration before clinical symptoms emerge would transform both diagnosis and clinical trial design. We evaluated five biomarker discovery frameworks -- cerebrospinal fluid (CSF) proteomics, plasma neurofilament and phospho-tau assays, multi-omic blood profiling, neuroimaging-derived digital biomarkers, and ML-integrated multi-modal panels -- across four neurodegenerative diseases in a prospective cohort of 4,286 participants (1,842 AD, 1,124 PD, 648 ALS, 672 FTD) from five European memory and movement disorder clinics (Vienna, Madrid, Berlin, Amsterdam, and London) followed for a median of 4.2 years (2016-2022). Our Neurodegenerative Biomarker Utility Score (NBUS) integrates diagnostic accuracy, preclinical detection capability, disease progression tracking, treatment response sensitivity, and clinical accessibility. ML-integrated multi-modal panels achieved the highest NBUS (0.928), detecting preclinical AD 6.4 years before symptom onset (AUROC 0.92) by combining plasma p-tau217, GFAP, NfL, and amyloid PET features. Plasma biomarkers alone (NBUS 0.904) achieved AUROC 0.90 for AD diagnosis -- approaching CSF performance while requiring only a blood draw.