Machine learning-based identification of biomarkers and mechanisms for the sepsis risk induced by aflatoxin B1 exposure

Su Shuncheng; Guo Jian; Xia Yichun; Wang Hanzhou; Li Yuhang; Ma Lele; Fu Yuqing; Hu Guanyu; Liu Zhu; Qian Yiming

doi:10.3969/j.issn.1002-1949.2026.03.006

Abstract：Objective This study aims to systematically investigate the molecular link between the environmental toxin aflatoxin B1(AFB1)and sepsis by integrating bioinformatics and machine learning approaches,identify key overlapping genes and core biomarkers,and explore the potential mechanisms by which AFB1 influences the progression of sepsis.Methods Sepsis-related transcriptomic datasets were obtained from the GEO database,merged,and subjected to batch effect correction before being divided into training and test sets.Sepsis-associated genes were screened through differential expression analysis and weighted gene co-expression network analysis(WGCNA),and potential targets of AFB1 were predicted from multiple databases.The intersection of these gene sets yielded overlapping genes.Gene ontology(GO)and kyoto encyclopedia of genes and genomes(KEGG)enrichment analyses were performed on the overlapping genes,and a protein-protein interaction(PPI)network was constructed to identify core genes.Multiple machine learning algorithms were employed to develop a diagnostic model,which was validated on an independent dataset.SHAP interpretability analysis was employed to identify the key biomarker with the highest contribution.Molecular docking was performed to assess the binding potential between the screened core biomarker and AFB1,with binding energy calculated to evaluate affinity.Results A total of 55 overlapping genes between AFB1 and sepsis were identified,which were significantly enriched in immune and inflammatory response pathways.The PPI network revealed 20 core genes,including EGFR,TP53,and SRC.The machine learning model(glmBoost+StepGlm[both]demonstrated excellent diagnostic performance(AUC=0.974).SHAP analysis identified GAPDH as the biomarker with the highest diagnostic value(AUC=0.901).Molecular docking confirmed a strong binding potential between AFB1 and the GAPDH protein(binding energy=-10.5 kcal/mol).Conclusions This study suggests that AFB1 may increase susceptibility to sepsis by directly targeting key molecules such as GAPDH,thereby disrupting immune homeostasis.GAPDH was identified as a valuable biomarker,providing new insights into the role of environmental toxins in sepsis pathogenesis and offering potential targets for risk assessment and early warning.