Integrative analysis of thiamethoxam induced hepatocellular carcinoma toxicity mechanisms

Abstract

Neonicotinoid (NEO) pesticides play a crucial role in agricultural production. However, their potential risks to human health and the environment cannot be overlooked. To gain a comprehensive understanding of the toxicity and mode of action of NEOs, thiamethoxam (THX), which exhibits the highest potential for carcinogenicity and hepatotoxicity, was selected as the subject of this study. We identified 61 intersection genes between THX targets and hepatocellular carcinoma (HCC)-related genes. These genes were then uploaded to the Metascape database for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses. The GO analysis indicated that the significant biological processes mainly involved the response to xenobiotic stimuli, cellular response to chemical stress, cellular response to biotic stimuli, and response to toxic substances. The KEGG enrichment analysis pinpointed several key pathways, primarily including the cell cycle and Glycolysis/Gluconeogenesis. Subsequently, the intersection genes were imported into the Gene Expression Profiling Interactive Analysis (GEPIA) and Gene Expression Omnibus (GEO) databases to analyze expression differences, leading to the identification of 15 significantly differentially expressed core genes (SDECGs). By applying the Support Vector Machine (SVM) machine-learning model, we screened out five feature genes (CYP2C19, CYP3A4, FBP1, THBS4, CYP7A1) and constructed a nomogram. Molecular docking of THX with these five feature genes showed binding energies of less than -5 kcal/mol. This study offers a theoretical foundation for understanding the underlying mechanisms of THX-induced HCC. The findings provide a scientific basis for the safety assessment of THX in agricultural applications and contribute to the establishment of pesticide safety standards.

Keywords: Hepatocellular carcinoma; Machine learning; Molecular docking; Neonicotinoid pesticide; Network toxicology; Thiamethoxam.

Fig. 1

Venn diagram of the intersection between HCC and THX target sets.

Fig. 2

GO and KEGG analysis of THX and HCC-Related Targets (A–C) Diagrams of BP, CC, and MF results in GO analysis; (D) KEGG analysis diagram.

Fig. 3

Differential gene expression profiling and interaction analysis. (A) Box plot showing the expression differences of core genes between normal and HCC samples; (B) Heatmap of SDECG expression in normal and HCC samples; (C) Circular plot depicting the chromosomal locations of core genes; (D) Correlation network of SDECGs; (E) Correlation analysis between the two SDECGs. Tip: ** is p < 0.01; *** is p < 0.001.

Fig. 4

Immune cell infiltration analysis of SDECGs in HCC. (A) Box plot comparing immune cell fractions between normal and HCC samples; (B) Heatmap of the correlation analysis between SDECGs and immune cells. Tip: * is p < 0.05; ** is p < 0.01; *** is p < 0.001.

Fig. 5

Machine learning for screening the key genes. (A) ROC curves for SVM, RF, XGB, GLM; (B) Box plots of residuals for SVM, RF, XGB, GLM; (C) Reverse cumulative distribution of residuals for SVM, RF, XGB, GLM machine learning models; (D) Bar plot showing the feature importance of SVM, RF, XGB, GLM machine learning models; (E) Nomogram was constructed based on five feature genes; (F) Calibration curve of the feature gene nomogram for THX-induced HCC; (G) Decision curve for the feature gene nomogram for THX-induced HCC; (H) ROC curve for the test GEO dataset; (I) ROC curves for external validation using the GEO dataset.

Fig. 6

The molecular docking results of THX with the top five genes. The molecular docking analysis of THX binding to CYP2C19 (A), CYP3A4 (B), FBP1 (C), THBS4 (D) and CYP7A1 (E) was conducted using AutoDock Vina software,