Construction of a pyroptosis-related gene-based diagnostic model for osteomyelitis and analysis of the mitochondria-inflammation interaction mechanism: a bioinformatics study integrating multi-omics and machine learning

Staphylococcus aureus (SA)-induced osteomyelitis (OM) is a common refractory orthopedic infection, presenting substantial challenges in early diagnosis and immune microenvironment characterization. Expression profiles of pyroptosis-related genes (PRGs) are closely associated with SA-OM; these genes may influence the immune microenvironment through mitochondrial-related pathways, thereby participating in disease progression, and specific gene combinations can be utilized to construct high-precision diagnostic models. This study aims to integrate multi-omics and machine learning to screen key pyroptosis-related diagnostic biomarkers in SA-OM, construct a high-precision diagnostic model, and elucidate its molecular mechanism influencing the immune microenvironment through “mitochondria-inflammation” crosstalk.

Based on 3 SA-OM datasets (GSE6269/GSE16129/GSE30119) retrieved from the GEO database, a total of 143 SA-OM patients and 79 healthy control samples were enrolled. Data preprocessing (batch effect correction using the sva package), differential expression analysis (DE-PRGs screened via the limma package, adj. P<0. 05 & |log₂ FC| >0. 263), co-expression network construction (key module genes identified through WGCNA algorithm, softThreshold=5), multi-omics cross-validation (Pearson correlation analysis for MR-PRGs screening), machine learning modeling (feature genes selected via SVM-RFE/LASSO/random forest cross-validation, n=9), and diagnostic model construction (logistic regression nomogram model, efficacy evaluated through AUC, calibration curve slope, and DCA) were performed, combined with immune microenvironment analysis (CIBERSORT/ssGSEA quantitative analysis of 22 immune cell infiltration levels).

Among 23 DE-PRGs, a diagnostic model comprising 8 key genes demonstrated excellent performance in both the training set (AUC=0. 89, 95%CI: 0. 83 to 0. 95) and validation set (AUC=0. 83, 95%CI: 0. 76 to 0. 90). RT-qPCR experiments further validated that the mRNA expression levels of the key pyroptosis pathway genes Caspase-1 and IL-18 in the SA-OM group were significantly upregulated compared with the control group (P<0. 05), corroborating the bioinformatics findings. The METTL3-MRPL39 axis was significantly enriched in “metabolic pathways” and “mitochondrial gene expression” biological processes. Furthermore, Th1/Th17 cell infiltration levels in the disease group were 3. 2-fold higher than those in the control group (P<0. 001), and METTL3 expression exhibited positive correlation with effector T cell infiltration (r=0. 65, P=0. 008).

This study systematically elucidates the regulatory network of pyroptosis-related genes in SA-OM. The constructed diagnostic model provides a novel tool for early screening, while the identified mitochondrial-inflammation interplay mechanisms and specific immune microenvironment characteristics establish a theoretical foundation for the development of targeted therapeutic strategies.