OBJECTIVE: This exploratory study attempted to identify a peripheral blood-based, ferroptosis-related gene signature and to construct a candidate ferroptosis driver scoring system that may assist the diagnosis of Parkinson's disease (PD) using bioinformatics and machine learning. METHODS: Peripheral blood gene expression profiles from PD patients and healthy controls were retrieved from the Gene Expression Omnibus (GEO). Differentially expressed genes (DEGs) between PD and controls were identified using the limma and intersected with ferroptosis-related genes curated in FerrDb. Key biomarkers were selected by intersecting the outputs of LASSO regression and Random Forest models. A ferroptosis driver score was then constructed using principal component analysis (PCA) of the selected genes. RESULTS: In the GSE99039 training dataset, 3507 DEGs were identified, of which 89 were related to ferroptosis. By integrating LASSO regression and random forest algorithms, six genes were selected to construct the ferroptosis driver score. In the training set, this score was significantly higher in patients (P < 0.001) and yielded an AUC of 0.656(95% CI: 0.605-0.707). This AUC outperformed single-gene classifiers (which ranged from 0.398 to 0.642), indicating its potential advantage as an auxiliary diagnostic tool. CONCLUSION: We constructed and preliminarily validated a peripheral blood-based ferroptosis driver scoring system that may serve as a candidate adjunctive tool for PD diagnosis. These findings are hypothesis-generating and may help guide future mechanistic and clinical studies on ferroptosis-related pathways in PD.