To evaluate the clinical potential of our neuronal cfDNA cell of origin classifier in identifying and distinguishing neurodegenerative disease.

Alzheimer’s disease (AD) is the most common neurodegenerative disorder, yet current diagnostics remain impractical for early, presymptomatic screening. As a result, diagnosis often occurs late in the disease course, when treatment options are limited. Blood-based assays of neuropathology and neurodegeneration, while less invasive, lack the regional and cell-type specificity needed for precise characterization to guide further diagnostic testing. 

Circulating cell-free DNA (cfDNA) released from dying neurons offers a promising alternative. cfDNA retains the methylation signatures of its cellular origin, providing cell-type-specific resolution of neurodegeneration with broad applicability to other diseases marked by selective neurodegeneration.

To evaluate clinical potential, we generated a methylation atlas of six neural cell types, including cortical, dopaminergic, and spinal motor neurons, and astrocytes, Schwann cells, and microglia, using native nanopore sequencing. Differentially methylated regions (≥10 CpGs, ≥10× coverage per CpG, ≥20% methylation difference, adj. p<0.05) were used to train classifiers that assign cfDNA fragments to their neural origin.

These classifiers were applied to 219 plasma samples from patients with AD, Parkinson’s disease (PD), amyotrophic lateral sclerosis (ALS), and controls.

Cortical neuron-derived cfDNA was significantly elevated in AD and mild cognitive impairment (MCI), distinguishing AD from controls, PD, and ALS (AUC = 1.000, 95% CI: 1.000–1.000). Dopaminergic neuron-derived cfDNA accurately identified PD (AUC = 0.985, 95% CI: 0.9688–1.000), while spinal motor neuron-derived cfDNA distinguished ALS (AUC = 0.991, 95% CI: 0.975–1.000). 

Together, these results highlight the potential of native cfDNA methylation profiling for noninvasive detection and differentiation of neurodegeneration in AD, PD, and ALS from a single blood sample. This scalable framework holds promise for earlier detection, longitudinal monitoring, therapeutic assessment, and application to other disorders of selective neurodegeneration.