Integrative Machine Learning and Experimental Validation Identify FIS1 as a Candidate Biomarker Linked to Mitochondrial Dynamics in Pulmonary Hypertension.

Pulmonary hypertension (PH) is characterized by progressive pulmonary vascular remodeling and a paucity of effective therapeutic interventions. Although dysregulated mitochondrial dynamics are implicated in this remodeling process, the key regulatory molecules and downstream mechanisms remain incompletely defined. This study aimed to systematically characterize molecular alterations associated with mitochondrial dynamics in PH and to explore the functional relevance and potential mechanisms of prioritized candidate genes. We integrated transcriptomic datasets from PH models with MitoCarta annotations to identify mitochondria-related differentially expressed genes. Candidate genes were prioritized using WGCNA and three machine-learning algorithms (LASSO, SVM-RFE, and random forest). These candidates were then experimentally evaluated in a hypoxia-induced PH mouse model and hypoxia-stimulated mouse pulmonary artery smooth muscle cells (mPASMCs) using qRT-PCR, Western blotting, immunohistochemistry, and transmission electron microscopy. Functional assays and assessments of mitochondrial injury were performed to investigate pathogenic relevance. Our analysis identified four key genes, with FIS1 showing high ROC/AUC-based discriminatory performance in both the training dataset and the independent replication dataset. Hypoxia was associated with increased FIS1 expression, mitochondrial fragmentation, loss of mitochondrial membrane potential, and ROS accumulation. We further observed that FIS1 knockdown suppressed mPASMC proliferation and migration, alleviated mitochondrial injury, and attenuated ferroptosis-associated alterations, accompanied by reduced lipid peroxidation, decreased Fe