INTEGRATED MULTI-OMICS AND MACHINE LEARNING APPROACHES REVEAL THE CHARACTERISTICS OF MACROPHAGES IN POSTMENOPAUSAL OSTEOPOROSIS AND IDENTIFY SMAD7 AS A KEY REGULATOR

Postmenopausal osteoporosis (OP) is a chronic bone disease characterized by low BMD and microstructural deterioration. While early research mainly focused on inhibiting osteoclast activity, recent perspectives highlight the interaction of bone resorption and formation. Given that macrophages serve as precursors to osteoclasts and secret cytokines that stimulate osteoblasts, they may be promising for identifying new dual-effect drug targets. In this study, we employed multi-omics and machine learning methods to investigate the molecular characteristics of macrophages in OP and identify some novel drug targets. We suggest that this research will help understand OP's molecular mechanisms and assist the development of new therapies. We integrated 6 scRNA-seq and 12 bulk RNA-seq samples as a discovery cohort to identify potential drug targets. Additionally, we utilized 20 paraffin tissue and 20 bulk RNA-seq samples for validation. Differential gene expression analysis was performed on scRNA-seq data, while Weighted Gene Co-expression Network Analysis (WGCNA) was utilized to analyze bulk RNA-seq data. Furthermore, we employed Lasso regression and Support Vector Machine (SVM) models to identify core genes. Subsequently, we performed Protein-Protein Interaction (PPI) analysis, cellular communication inference, differentiation trajectory analysis, and network pharmacology analysis based on the identified genes. In discovery cohort, we observed a significant difference in the proportion of macrophages (Fig.1A) . We identified modules correlated with macrophages and detected the differentially expressed genes (Fig.1B-C). The KEGG enrichment analysis revealed that these differential genes were significantly enriched in ferroptosis-related pathways (Fig.1D) . Additionally, the core genes identified in PPI network showed a significant correlation with ferroptosis ( Figure 2A-B ). Moreover, machine learning techniques identified 4 core genes, with Smad7 emerging as the most critical, showing a strong connection in validation cohort ( Figure 2C-D ). Finally, we validated the expression of Smad7 and the differential proportions of macrophages in paraffin tissues. In conclusion, we detected the significantly different proportion of macrophages in osteoporosis, along with identifying differentially expressed genes. Enrichment analysis and PPI network indicated these genes were highly correlated with ferroptosis mechanism, which may help explain the varying distribution of macrophages. Machine learning approaches further identified several novel targets, with Smad7 emerging as the most significant and promising regulator in validation cohort. Our findings provide a novel insight for OP targeted treatments, while also contributing to the understanding of underlying molecular mechanisms. For any figures or tables, please contact the authors directly.