Abstract
BACKGROUND: Dementia with Lewy bodies (DLB) exhibits a more aggressive progression and poorer prognosis than Alzheimer's disease (AD), yet clinical differentiation remains challenging. Dysregulated lipid metabolism, implicated in α-synuclein aggregation and neuroinflammation, may offer specific biomarkers for distinguishing DLB and AD. METHODS: This cross-sectional study implemented targeted lipidomic profiling to comprehensively characterize plasma lipidomes in a cohort comprising 50 DLB patients and 56 AD patients. Five machine learning algorithms - least absolute shrinkage and selection operator (LASSO) regression, support vector machine (SVM), random forest (RF), recursive feature elimination (RFE), and stepwise regression - were systematically applied for biomarker discovery. RESULTS: Significant alterations were observed in 7 lipid classes and 65 specific lipid species in DLB compared to AD. DLB plasma exhibited marked elevations in sphingolipids (total Cer, Hex1Cer, SM), lysophospholipids (LPC, LPE), phosphatidic acid (PA), alongside significant reductions in 45 triacylglycerol (TG) species compared to AD. Five machine learning algorithms consistently identified PA(16:0_16:0) and PA(16:0_20:4) as core discriminators between DLB and AD. The LASSO regression model demonstrated superior generalizability in the test set (AUC=0.916), selecting a 11-lipid panel dominated by PA species, alongside PC(18:0_20:4), ChE(22:4), Hex2Cer(d18:1_22:0), and PE species. CONCLUSION: This first comprehensive targeted lipidomics study reveals distinct plasma lipid signatures differentiating DLB from AD, characterized by upregulated sphingolipids, lysophospholipids, and PA, and downregulated TG. Machine learning identified a 11-lipid biomarker panel, highlighting profound disturbances in glycerophospholipid metabolism. These findings provide novel molecular insights into DLB pathogenesis and a promising diagnostic tool for diagnosis.