Network toxicology integrated with single-cell RNA sequencing reveals ROS-related targets of paraquat in pulmonary fibrosis

Objective This study explored the cellular heterogeneity and molecular targets driving the transition from acute injury to paraquat (PQ)-induced pulmonary fibrosis (PF) through a multi-omic and single-cell approach. Methods Chemical–gene interactions for paraquat (PQ) and pulmonary fibrosis (PF) gene–disease associations were retrieved from CTD and standardized, with PF genes restricted to CTD “Direct Evidence” records to improve confidence. The resulting PQ–PF candidates were mapped to STRING for PPI topology, and to a paraquat-injured human lung scRNA-seq dataset (GSE231647) for cell-type localization and communication inference (Seurat v5.0.1; cellchat version 2.1). A ROS–Antioxidant–Transporter prioritization scheme was used to nominate redox-related targets, followed by blind molecular docking (CB-Dock2) for structural plausibility. Results Direct-evidence filtering yielded 122 PF genes, and intersection with 4,849 PQ targets produced an 82-gene PQ–PF nexus. PPI analysis identified inflammatory–fibrotic hubs (IL1B, TNF, IL6, TGFB1, STAT3, FN1, and MMP9) and remained robust under a higher STRING confidence threshold (≥0.7). Single-cell mapping localized ROS-related candidates to macrophages, airway epithelium, and myofibroblasts; CellChat highlighted prominent macrophage→fibroblast signaling via ECM remodeling and SPP1/MIF pathways. Docking supported structurally plausible micromolar-range interactions between PQ and antioxidant/stress-response proteins (GPX3, SOD1, and HMOX1; −6.67 to −5.93 kcal/mol). Conclusion Together, this network-to-single-cell-to-structure framework supports an oxidative-stress-centered, multi-target mechanism in PQ-induced lung fibrosis and nominates a GPX3/SOD1-centered antioxidant module and the inducible HMOX1 stress-response axis, along with upstream inflammatory–fibrotic hubs, as candidate intervention nodes for further experimental validation. The docking results are consistent with structurally plausible low-affinity PQ engagement, while remaining compatible with indirect redox-mediated modulation in vivo.