Introduction:

Current diagnostic techniques cannot reliably detect neurodegeneration before at the earliest stages of Alzheimer’s disease (AD).1 We aimed to develop a sequencing-based assay that utilises multiple features of cell-free DNA (cfDNA) in plasma as biomarkers of AD. The software developed, named NEON (NEuron-derived cfDNA biomarker of neurodegeneratiON), leverages DNA methylation (5mC), DNA hydroxymethylation (5hmC), nucleosome positioning (Windowed Positioning Score; WPS), and fragment end motif derived signals from a single sequencing experiment to detect the cfDNA-associated signatures of neuronal damage. 2,3

 

Results:
Figure 1: Deconvolution of cell type proportions  in plasma samples from Alzheimer’s disease patients & age-matched controls. Using NEONs cfDNA-optimised deconvolution algorithm, oligodendrocyte-derived cfDNA was found significantly be enriched in AD plasma (log2FC = 1.55, p = 0.026), whereas granulocyte-derived cfDNA was found to be significantly depleted in AD plasma (log2FC =   -0.34, p = 0.036).  Borderline significant enrichments were also observed in alveolar epithelium-derived cfDNA (log2FC = 1.46, p = 0.055) and neuron-derived cfDNA (log2FC = 0.62, p = 0.087) in AD plasma.

Figure 2: WPS and end motif analysis reveal characteristic fragmentomic differences in AD plasma. WPS analysis detects AD-associated open chromatin signatures at via plasma cfDNA, indicated by increased trace phasing. Fragment end motif analysis (bottom) suggests AD cfDNA is enriched for A-starting end motifs, suggesting decreased DNASE1L3-driven apoptotic fragmentation toward alternative nuclease activity.

Figure 3: Leave-One-Out cross validation using a deconvolution, window positioning score, fragment end motif, and a combined genomic biomarker feature panel. ROC analysis demonstrates  that a ridge regression model trained on a combined panel of the most discriminative genomic biomarker features achieves superior classification (AUC = 0.816) compared to deconvolution (AUC = 0.623), fragment end motif  (AUC = 0.707), and windows positioning core (AUC = 0.774) features alone.

 

Discussion:
Preliminary results for using NEON to detect neuron-derived cfDNA as a biomarker neurodegeneration are promising, however testing of diagnostic capability in full-scale patient populations as well as integration with further undiscussed genomic biomarkers are required to refine this approach before it is fit for use in aiding Alzheimer’s disease diagnosis