Integrative ferroptosis-related bioinformatics identifies PPARγ as a therapeutic target of ursolic acid against inflammation and pathological osteogenesis in ankylosing spondylitis

Background Ankylosing spondylitis (AS) is an immune-mediated inflammatory disease characterized by persistent inflammation and aberrant new bone formation. Ferroptosis has increasingly been implicated in immune dysregulation and pathological bone remodeling; however, its molecular involvement in AS and its potential value for natural product discovery remain poorly understood. This study aimed to identify ferroptosis-related molecular targets in AS, screen natural compounds targeting the key candidate, and validate their therapeutic effects and underlying mechanisms. Methods Bulk RNA-seq datasets were combined to perform differential expression profiling, ferroptosis evaluation, weighted co-expression network construction, functional pathway annotation, and machine-learning-driven feature screening. Subsequently, single-cell transcriptomic data were examined to map the cell-type-specific localization of the hub genes that had been pinpointed. Natural product and compound–target databases were subsequently searched for small-molecule compounds associated with both AS and the key target. Candidate compounds were subjected to batch molecular docking. Based on its predicted binding properties and biological relevance, Ursolic acid (UA) was selected as a candidate drug. Molecular dynamics simulations were performed to evaluate the stability of the UA–PPARγ complex. The therapeutic effects of UA were subsequently examined in an AS mouse model and in BMP-2/4-induced osteogenic differentiation of fibroblasts. Histological staining, qRT-PCR, and western blotting were used to assess inflammatory responses, pathological osteogenesis, and PPARγ-related signaling. Results Integrated transcriptomic analysis identified 20 genes shared among AS-related differentially expressed genes, ferroptosis-related genes, and genes from ferroptosis-associated co-expression modules. Machine-learning analysis further identified PHF21A, PPARG , and TP53INP1 as hub genes, among which PPARG showed the highest predictive contribution and distinct expression patterns across immune cell populations. Database-based natural product screening and batch molecular docking identified UA as a promising PPARγ-targeting candidate. Molecular docking and molecular dynamics simulations supported a potential interaction between UA and PPARγ. In vivo , UA markedly alleviated inflammatory infiltration, reduced the expression of TNF-α, IL-6, and IL-17, and attenuated pathological bone formation in AS mice. In vitro , UA increased PPARγ expression while suppressing the osteogenic markers COL1A1, ALP, RUNX2, OPN, and BMP-2, thereby inhibiting the osteogenic transdifferentiation of fibroblasts. Conclusion This study established an integrated strategy combining ferroptosis-related bioinformatics analysis, machine-learning-based target identification, natural product screening, molecular simulation, and experimental validation. PPARγ was identified as a ferroptosis-associated molecular target in AS, and UA exerted concurrent anti-inflammatory and anti-osteogenic effects accompanied by increased PPARγ protein expression. These findings suggest that targeting PPARγ with UA might be a potential therapeutic strategy for AS.