Combining network toxicology with machine learning and experiment validation to analyze the molecular mechanism and core target screening of aristolochic acid nephropathy

Abstract This study aimed to comprehensively investigate the possible toxicity and molecular mechanisms of aristolochic acid (AA)-induced aristolochic acid nephropathy (AAN) and to provide a scientific basis for its prevention and treatment. Using network toxicology, machine learning, multidimensional bioinformatics, molecular docking, and molecular dynamics simulation, the study first collected AA targets and AAN-related targets from multiple databases. After integration and deduplication, 290 intersecting targets were identified. Functional pathways were then explored through gene ontology (GO) and Kyoto Encyclopedia of genes and genomes (KEGG) enrichment analysis. Least absolute shrinkage and selection operator (LASSO) regression and support vector machines (SVM) algorithms were used to pinpoint core targets. The expression and diagnostic relevance of these core targets were verified using Gene Expression Omnibus (GEO) datasets. Immune infiltration analysis was performed, and molecular docking and dynamics simulations were performed to construct adverse outcome pathways (AOP). Finally, The expression of core genes was measured in HK2 cells after 48 hours of intervention with AA. Results showed that the intersectional targets were primarily enriched in lipid metabolism and related pathways. Three core targets, solute carrier family 1 member 3 (SLC1A3), CAMP-specific phosphodiesterase 4B (PDE4B), and fatty acid binding protein 3 (FABP3), were identified, all with area under the curve (AUC) values > 0.9, demonstrating high diagnostic specificity and association with multiple immune cell infiltrations. Molecular docking revealed that AA binding energies to all three core targets were