Dementia etiology classification using NULISA plasma biomarkers and machine learning

Area of Science:

• Neuroscience
• Biomarker Discovery
• Computational Biology

Background:

• Accurate antemortem diagnosis of dementia subtypes is challenging, especially with atypical presentations.
• Multiplexed plasma proteomics offers a minimally invasive diagnostic approach.
• Machine learning enhances the analysis of complex proteomic data.

Purpose of the Study:

• To evaluate the feasibility of using multiplexed plasma proteomics with machine learning for differential diagnosis of dementia etiologies.
• To identify disease-specific protein patterns in plasma for various dementia types.
• To develop and validate machine learning models for predicting dementia etiology.

Main Methods:

• Plasma samples from 194 participants were analyzed using the NULISA CNS 120+ biomarker panel.
• Differentially abundant proteins associated with Alzheimer's disease, frontotemporal lobar degeneration, Lewy body disease, and vascular disease were identified.
• Supervised XGBoost classifier models were trained and applied to predict etiology in participants with mild cognitive impairment.

Main Results:

• NULISA plasma biomarkers revealed distinct protein patterns specific to each dementia etiology.
• XGBoost classifiers accurately differentiated between the dementia subtypes with high specificity.
• The models successfully generated robust etiologic predictions when applied to mild cognitive impairment cases.

Conclusions:

• Multiplexed plasma proteomics combined with machine learning is a feasible approach for diagnosing complex dementia etiologies.
• This framework provides a data-driven method for differential diagnosis of neurodegenerative diseases.
• Future research with larger cohorts will further refine biomarker identification and diagnostic accuracy.