Identification of lipid metabolism-related biomarkers in familial hypercholesterolemia via integrated bioinformatics and machine learning approaches

Familial hypercholesterolemia (FH) is a genetic disorder characterized by imbalances in lipid metabolism, markedly increasing cardiovascular risk. Identifying lipid metabolism-related biomarkers is essential for understanding FH pathogenesis and developing therapeutic strategies. A comprehensive bioinformatics analysis was conducted using 776 lipid metabolism-related genes (LMRGs) from the MSigDB database and the GSE6054 dataset containing 10 FH and 13 healthy samples. Differentially expressed genes (DEGs) intersected with LMRGs to obtain candidate genes, followed by random forest and SVM-RFE machine learning to identify key biomarkers. Functional enrichment, subcellular localization, co-expression, transcription factor (TF) prediction, ceRNA network, and drug screening were subsequently performed. We identified 429 DEGs and 13 candidate genes, refined to six key genes (STAR, GRHL1, BZRAP1, LGMN, PLA2G4D, PLA2G12B). ROC analysis demonstrated that all six key genes exhibited AUC values greater than 0.7, underscoring their diagnostic potential. Functional enrichment further revealed significant associations with the ribosome and spliceosome pathways. Subcellular localization suggested mitochondrial and extracellular functions. Co-expression showed a significant positive correlation between GRHL1 and PLA2G12B. TF prediction revealed 17 TF-gene interactions, while ceRNA analysis outlined regulatory relationships. Drug prediction identified 224 potential therapeutic compounds, with STAR showing the most interactions. This study highlights six lipid metabolism-related biomarkers in FH through integrated bioinformatics and machine learning. These findings provide new insights into FH mechanisms and potential therapeutic targets.