Investigating the Association Between Azo Dyes and Skin Cancer Using Network Toxicology, Machine Learning, and Molecular Docking

Background: In recent years, azo dyes—widely used in textiles, hair dyes, paints, plastics, and rubber for coloration—have been demonstrated to associate with various diseases, particularly carcinogenesis. This study aims to investigate the potential association between azo dyes and skin cancer, and to identify the pivotal genes and pathways potentially involved in this process. Methods： This study utilized multiple online databases to identify target genes associated with azo dyes and skin cancer. Protein-protein interaction (PPI) analysis and visualization were performed on the intersecting genes, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses to elucidate potential mechanisms. Subsequently, an optimal predictive model was selected by combining eight machine learning algorithms. Prognostic models were constructed using Gradient Boosting Machine (GBM) and Support Vector Machine (SVM), and their performance was validated across four independent external datasets. Finally, molecular docking analysis was conducted to explore interactions between key genes and azo dyes. Results： Seven common azo dyes were selected for analysis: Benzidine, 4-Aminobiphenyl, 4-Nitrobiphenyl, 2-Naphthylamine, o-Aminoazotoluene, 2,4-Diaminotoluene, and 2,4,5-Trimethylaniline. A total of 164 cross-targets associated with skin cancer were identified. GO and KEGG enrichment analyses revealed that these targets primarily participate in carcinogenesis and proliferation-related biological processes through pathways involving metabolism, cancer, inflammation, as well as key signaling pathways such as MAPK and PI3K-AKT.Comparative evaluation of eight machine learning models identified GBM (Gradient Boosting Machine) as the optimal predictive model. The prognostic model, constructed by integrating GBM with SVM (Support Vector Machine), highlighted six key genes associated with both azo dyes and skin cancer. This model demonstrated robust predictive performance across three independent external validation datasets. Furthermore, molecular docking analysis confirmed potential interactions between these azo dyes and the core targets. Conclusion： Research findings demonstrate that TNF, MAPK8, MMP9, GSK3B, CDC42, and SIRT1 play pivotal roles in the process of azo dye-induced skin cancer pathogenesis. These results provide novel insights into the molecular mechanisms linking azo dye exposure to skin carcinogenesis, while highlighting the imperative need for enhanced public awareness regarding safety concerns associated with textile dyes, hair coloring products, and related consumer goods.