Work

Machine learning combined with multidimensional validation identifies SNCA and NR2F2 as key hubs implicated in paraquat and rotenone co-exposure-induced Parkinson’s disease

Year: 2026

Type: article

Abstract: Combined exposure to pesticides, such as paraquat (PQ) and rotenone (Rot), is a significant risk factor for Parkinson’s disease (PD). However, the synergistic pathogenic mechanisms underlying their co-exposure remain largely elusive. This study aimed to elucidate the key molecular hubs and interaction networks driving PD pathogenesis following combined PQ and Rot exposure. An integrated analytical framework combining bioinformatics, machine learning (ML), and multi-faceted validation was employed. Pesticide targets were retrieved from toxicology databases and intersected with PD-associated genes identified from GEO transcriptomic data via differential expression analysis and WGCNA. Candidate targets underwent functional enrichment analysis. A suite of ML algorithms, coupled with SHAP analysis, was utilized to identify and rank core genes. Finally, molecular docking, molecular dynamics (MD) simulations, and in vitro experiments were used for validation. We identified 41 common pesticide-disease targets, predominantly enriched in neurotransmitter receptor signaling and dopaminergic synaptic pathways. ML analysis pinpointed four core genes (SNCA, NR2F2, ARG1, and HTR2A), with SNCA and NR2F2 identified as the most significant predictors. Molecular docking and MD simulations indicated stable binding interactions between the pesticides and these core proteins. Furthermore, in vitro co-treatment with PQ and Rot synergistically upregulated SNCA and NR2F2 expression. PQ and Rot co-target specific genes and signaling pathways to induce synergistic neurotoxicity, thereby promoting PD pathology. SNCA and NR2F2 are identified as key hubs implicated in this synergy. These findings provide novel insights into the mechanisms of co-exposure and highlight potential biomarkers for early detection and therapeutic intervention.

Source: Human Genomics

DOI: 10.1186/s40246-026-01048-0