Abstract
Abstract: Background
Lewy body dementias (LBD), comprising dementia with Lewy bodies (DLB) and Parkinson’s disease dementia (PDD), are defined by misfolded α-synuclein aggregation. Seed amplification assays (SAAs), such as RT-QuIC, enable sensitive detection of α-synuclein aggregates but typically provide binary readouts and require fluorescence labeling. Raman spectroscopy offers a label-free approach to detect subtle biochemical changes, and its diagnostic potential can be enhanced with machine learning.
Objectives
This proof-of-concept study aimed to evaluate whether Raman spectroscopy combined with machine learning can improve SAA-based discrimination of LBD from controls in cerebrospinal fluid (CSF).
Methods
We analyzed a small number of post-mortem CSF samples from pathologically confirmed DLB (n = 2), PDD (n = 2), and controls (n = 2) using a 7-day SAA. Raman spectra were collected on Days 1, 4, and 7 and analyzed using principal component analysis (PCA) and uniform manifold approximation and projection (UMAP).
Results
Following SAA, both PCA and UMAP distinguished combined LBD samples from controls within 24 h (Day 1), reflecting biochemical changes consistent with α-synuclein fibrillation. Spectral shifts indicated decreased α-helical content with increased β-sheet structures. No consistent separation between DLB and PDD was observed.
Conclusion
This preliminary study demonstrates that combining Raman spectroscopy with machine learning can enable rapid, label-free detection of disease-specific changes. Despite the very limited sample size, these findings highlight the potential of this novel workflow and strongly warrant its validation in larger cohorts.
Keywords: Lewy body dementia; Raman spectroscopy; α-synuclein aggregation; Machine learning analysis; Diagnostics
Lewy body dementias (LBD), comprising dementia with Lewy bodies (DLB) and Parkinson’s disease dementia (PDD), are defined by misfolded α-synuclein aggregation. Seed amplification assays (SAAs), such as RT-QuIC, enable sensitive detection of α-synuclein aggregates but typically provide binary readouts and require fluorescence labeling. Raman spectroscopy offers a label-free approach to detect subtle biochemical changes, and its diagnostic potential can be enhanced with machine learning.
Objectives
This proof-of-concept study aimed to evaluate whether Raman spectroscopy combined with machine learning can improve SAA-based discrimination of LBD from controls in cerebrospinal fluid (CSF).
Methods
We analyzed a small number of post-mortem CSF samples from pathologically confirmed DLB (n = 2), PDD (n = 2), and controls (n = 2) using a 7-day SAA. Raman spectra were collected on Days 1, 4, and 7 and analyzed using principal component analysis (PCA) and uniform manifold approximation and projection (UMAP).
Results
Following SAA, both PCA and UMAP distinguished combined LBD samples from controls within 24 h (Day 1), reflecting biochemical changes consistent with α-synuclein fibrillation. Spectral shifts indicated decreased α-helical content with increased β-sheet structures. No consistent separation between DLB and PDD was observed.
Conclusion
This preliminary study demonstrates that combining Raman spectroscopy with machine learning can enable rapid, label-free detection of disease-specific changes. Despite the very limited sample size, these findings highlight the potential of this novel workflow and strongly warrant its validation in larger cohorts.
Keywords: Lewy body dementia; Raman spectroscopy; α-synuclein aggregation; Machine learning analysis; Diagnostics
| Original language | English |
|---|---|
| Journal | Journal of Neuroscience Methods |
| Volume | 425 |
| Early online date | 14 Nov 2025 |
| Publication status | E-pub ahead of print - 14 Nov 2025 |