A comparison of machine learning classifiers for dementia with Lewy bodies using miRNA expression data

Area of Science:

• Neuroscience
• Biomarker Discovery
• Computational Biology

Background:

• Dementia with Lewy bodies (DLB) is a common neurodegenerative disease, second only to Alzheimer's disease.
• Current diagnostic methods for DLB have limitations in sensitivity.
• MicroRNAs (miRNAs) are emerging as promising biomarkers for early detection.

Purpose of the Study:

• To identify potential serum miRNA biomarkers for prodromal Dementia with Lewy bodies (DLB).
• To construct an accurate risk prediction model for DLB using machine learning.
• To investigate the association of miRNA target genes with DLB pathology.

Main Methods:

• Serum miRNA expression data from 478 Japanese individuals were analyzed.
• Machine learning algorithms including penalized regression, random forest, support vector machine, and gradient boosting decision tree were employed.
• A risk prediction model was built using 180 miRNAs and two clinical features.

Main Results:

• The gradient boosting decision tree model achieved an accuracy of 0.829 on an independent test set.
• Gene set enrichment analysis identified 6 functional genes in the DHA signaling pathway linked to DLB.
• BCL2L1 and PIK3R2 showed statistically significant associations with DLB through gene-based association studies.

Conclusions:

• The developed prediction model serves as an effective tool for DLB classification.
• BCL2L1 and PIK3R2 are identified as statistically significant genetic risk factors for DLB.