Identification and Experimental Validation of Candidate Biomarkers for Parkinson’s Disease through Integrated Bioinformatics, Mendelian Randomization, and Machine Learning

Background Identifying reliable biomarkers and therapeutic targets is important for improving the early diagnosis and treatment of Parkinson’s disease (PD). This study aimed to identify PD-related core genes, validate their expression, and explore potential therapeutic drugs. Methods PD transcriptomic datasets were analyzed using differential expression analysis, Mendelian randomization, and machine learning. Candidate genes were validated by qPCR in an MPTP-induced PD mouse model. Potential drugs were screened using the DSigDB database, followed by molecular docking and 100 ns molecular dynamics simulations to evaluate protein–ligand binding and stability. Results Seven core genes were identified, including AMIGO2, ATG7, CDK2AP2, CYP27A1, MT1X, NIPSNAP1, and TTC19. AMIGO2, ATG7, NIPSNAP1, and TTC19 were downregulated, whereas CDK2AP2, CYP27A1, and MT1X were upregulated. Animal experiments validated the abnormal expression of ATG7, NIPSNAP1, TTC19, CYP27A1, and MT1X. Drug screening identified 239 candidate compounds, of which five met the screening criteria. Molecular docking and molecular dynamics simulations indicated stable interactions between the top three drug–target complexes. Conclusion This study identified and experimentally validated several PD-related core genes and potential therapeutic compounds, providing candidate biomarkers and targets for further investigation of PD mechanisms and treatment.