Comprehensive Analysis of the Role of Metabolic Features in Osteoporosis: A Multi-Omics Analysis

Abstract

Purpose: This study aims to comprehensively explore the metabolic features related to the pathogenesis of osteoporosis (OP) through multi-omics analysis strategy.

Patients and methods: Gene expression profiles of OP patients (GSE56815) were downloaded from GEO, and metabolism-related genes (MRGs) were extracted. Plasma samples from 45 OP patients and 18 healthy controls (CON) were collected for metabolomics. We predicted miRNA and transcription factors (TFs) regulating the expression of MRGs on public databases ENCORI and JASPAR, and analyzed the expression levels of target miRNAs using miRNA sequencing of femoral tissues from 7 samples (OP:CON=4:3). Three machine learning algorithms were used to evaluate the diagnostic potential of metabolic signatures for OP.

Results: A total of 402 significantly differentially expressed MRGs (DEMRGs) were identified in the transcriptome, and these DEMRGs were enriched in 11 metabolic pathways (P<0.05). Metabolomics identified 119 differential plasma metabolites, enriched in 5 metabolic pathways (P<0.05). Purine metabolism, Tryptophan metabolism, and Tyrosine metabolism were identified as key metabolic pathways and were significantly enriched in DEMRGs. Femoral miRNA sequencing found 124 differentially expressed miRNAs, with 23 regulating key metabolic pathway gene expression (P<0.05). Additionally, 13 differentially expressed TFs were predicted to regulate the expression levels of these 23 miRNAs. Finally, three MRGs and one plasma metabolite were selected based on the machine learning algorithm, with AUC of 0.782, 0.714, 0.772 and 0.836, respectively. The diagnostic performance of these metabolic features was better than that of traditional bone metabolism biochemical markers.

Conclusion: This multi-omics study comprehensively explores the metabolic landscape in OP progression, highlighting the central role of metabolic features in the disease. The constructed multi-omics regulatory network aids in understanding the molecular mechanisms of metabolic features in OP progression.

Keywords: biomarkers; machine learning; metabolic features; multi-omics analysis; osteoporosis.

Figure 1

Comprehensive analysis of metabolic characteristics in the OP population. (A) Volcano plot of differentially expressed MRGs. Red indicated significantly up-regulated genes, and blue indicated significantly down-regulated genes. (B) GO enrichment analysis of differentially expressed MRGs. (C) The top 30 KEGG significantly enriched pathways. Metabolic pathways were highlighted in red font. (D) Volcano plot of differentially expressed plasma metabolites. Red represented significantly up-regulated metabolites, and blue represented significantly down-regulated metabolites. (E) KEGG enrichment analysis of significantly differential plasma metabolites. (F) Spearman correlation heatmap of metabolites from the 5 significantly enriched metabolic pathways with BMD T-scores. The color bar indicates the magnitude of the correlation, with red representing positive correlation and blue representing negative correlation. *P<0.05, **P<0.01.

Figure 2

Identification of miRNAs and TFs associated with OP metabolic pathways. (A) Volcano plot of differentially expressed miRNAs in bone tissue. miRNAs with P<0.05 and |log2(FC)|>1 were defined as significantly differential miRNAs. Red represented significantly up-regulated miRNAs, and green represented significantly down-regulated miRNAs. (B) Flowchart illustrating the miRNA prediction process. miRNAs predicted by at least two out of seven databases were retained. (C) VENN diagram showing the overlap of 23 miRNAs between significantly differential miRNAs and predicted miRNAs. (D) Identification of upstream TFs of miRNAs in the GSE56815 dataset. *FDR<0.05, **FDR<0.01.

Figure 3

Construction of the TF-miRNA-MRGs-metabolite network. Target regulatory relationships between TFs, miRNAs, and mRNAs were established based on database predictions. mRNAs and metabolites were co-involved in the alteration of metabolic pathways, and mRNAs and metabolites enriched in different metabolic pathways show significant correlations. Triangles represented TFs, circles represented metabolic mRNAs, squares represented plasma metabolites, and diamonds represented miRNAs. Solid lines indicated significant correlations (Spearman correlation), and dashed lines indicated target regulatory relationships. Red font indicated up-regulated expression, and blue font indicated down-regulated expression.

Figure 4

Machine learning screening for metabolic features biomarkers. (A) LASSO logistic regression coefficient penalty plot. (B) Model validation using AUC values with RF. (C) SVM cross-validation. (D) Three machine learning algorithms share 3 MRGs. (E) ROC analysis of MRGs. (F) ROC analysis of the plasma metabolite N-Acetyl-S-(N-methylcarbamoyl)-cysteine shared by three machine learning algorithms. *FDR<0.05, **FDR<0.01.