**Background:** Meibomian gland dysfunction (MGD) is a common ocular disorder causing evaporative dry eye, with prevalence estimates of 39–50% in the US population. Aging is a major risk factor, but the molecular mechanisms, particularly lipid metabolic alterations, remain poorly understood. This study aimed to comprehensively investigate age-related lipid metabolic changes in MGD using integrated multi-omics and machine learning (ML) to identify potential diagnostic and therapeutic targets.
**Methods:** Meibomian glands (MGs) were dissected from young (2-month, n=9) and aged (2-year, n=9) female C57/BL6J mice. Clinical examinations (slit-lamp, fluorescein staining, corneal opacity scoring) and Oil Red O staining confirmed MGD phenotype. For proteomics (n=3 per group), TMT-labeled LC-MS/MS was performed, identifying 5564 protein groups and 42,010 peptides. Differentially expressed proteins (DEPs) were defined by fold change >1.2 or <0.83 and p<0.05. Gene Ontology (GO) and KEGG pathway enrichment analyses were conducted. For lipidomics (n=6 per group), untargeted LC-MS/MS identified 1089 lipid species from 25 classes. Differentially abundant lipids were selected by VIP>1 and p<0.05. ML using LASSO regression and t-test was applied to construct a predictive lipid signature. Protein-protein interaction (PPI) networks were built using STRING.
**Key Results:** Aged mice showed significantly higher body weight (29.67±3.76 g vs. 16.73±0.66 g), increased corneal opacity and fluorescein staining scores, and atrophic MGs with decreased lipid droplets on Oil Red O staining. Proteomics revealed 375 DEPs (170 upregulated, 205 downregulated). Downregulated proteins were enriched in lipid metabolic process, cholesterol biosynthetic process, fatty acid beta-oxidation, and sterol biosynthetic process. KEGG analysis highlighted cholesterol metabolism and biosynthesis of unsaturated fatty acids. PPI networks identified key cholesterol biosynthesis proteins: Sqle, Fdft1, Lss, Dhcr24, and Msmo1. Lipidomics showed significantly lower total lipids in aged mice. The top five lipid subclasses in aged mice were ChE (56.877%), PC (12.644%), TG (9.912%), PE (9.199%), and PS (4.178%); in young mice: ChE (78.086%), PC (7.331%), PE (4.861%), TG (3.622%), and PS (2.236%). ChE and PIP significantly decreased, while TG and DGDG significantly increased in aged mice. Saturation ratios decreased for ChE, TG, and PS, but increased for PT and LPC. OPLS-DA showed excellent separation (R²X=0.847, R²Y=0.992, Q²=0.965). 47 lipid species (7 classes) were differentially expressed (VIP>1, p<0.05), including six downregulated ChE species: ChE(25:0)+NH4, ChE(26:0)+NH4, ChE(26:1)+NH4, ChE(28:1)+NH4, ChE(30:1)+NH4, and ChE(32:2)+NH4. ML LASSO regression identified a 9-lipid diagnostic signature: PE(36:1p)-H, PC(16:0/22:5)+HCOO, ChE(26:1)+NH4, ChE(26:0)+NH4, ChE(30:1)+NH4, TG(27:0/16:0/18:1)+NH4, TG(26:0/18:1/18:1)+NH4, PI(16:1/14:1)-H, and PC(36:4e)+HCOO. A hypothetical network was proposed linking cholesterol biosynthesis proteins (Dhcr24, Lss, Fdft1) and ChE molecules (ChE(26:0), ChE(26:1), ChE(30:1)) in ARMGD pathogenesis.
**Clinical Implications:** This study provides the first integrated multi-omics and ML analysis of age-related MGD, identifying a 9-lipid diagnostic signature and key protein biomarkers (Dhcr24, Lss, Fdft1) from cholesterol biosynthesis pathways. The findings suggest that decreased ChE (especially ChE(26:0), ChE(26:1), ChE(30:1)) and altered saturation patterns contribute to tear film instability and increased meibum viscosity, hallmark features of MGD. These biomarkers offer potential for developing novel diagnostic tools and targeted therapies for age-related MGD, though validation in human samples and functional studies are needed.