Deciphering the potential molecular links between ochratoxin A and colorectal cancer: an integrated computational toxicological and single-cell transcriptomic study.

BACKGROUND: Ochratoxin A (OTA), a common foodborne mycotoxin, is classified as a potential human carcinogen. However, the specific molecular mechanisms linking OTA exposure to colorectal cancer (CRC) pathogenesis remain poorly understood. METHODS: We employed an integrative computational approach combining network toxicology, multi-cohort transcriptomic analysis, and machine learning to identify potential connections between OTA and CRC. Functional enrichment analysis was conducted to characterize the associated biological processes. Molecular docking and molecular dynamics simulations were used to computationally assess the predicted binding potential and conformational stability of OTA-target complexes. Immune cell infiltration patterns were assessed using CIBERSORT, and protein-level expression patterns were examined using publicly available Human Protein Atlas (HPA) data. Single-cell RNA-sequencing (scRNA-seq) data from three public cohorts were integrated to characterize cell-type-specific expression patterns and to explore intercellular communication via CellChat. scTenifoldKnk-based in silico perturbation analysis of MET and PLAU was performed in their respective high-expression cell populations to generate functional hypotheses. RESULTS: Intersection of OTA-predicted targets with CRC-associated genes identified 81 candidate genes enriched in metabolic and microenvironment remodeling pathways. Integration of machine learning frequency, SHAP importance, and PPI topological centrality pinpointed five core genes: MET, PLAU, TOP2A, FABP4, and ADH1B. Molecular docking and dynamics simulations suggested favorable predicted binding and conformational stability for selected OTA-target complexes. Immune infiltration analysis revealed a myeloid-enriched microenvironment, with MET, PLAU, and TOP2A positively correlating with pro-tumorigenic subsets, while FABP4 and ADH1B were associated with lymphoid populations. HPA immunohistochemistry data were consistent with the upregulation of MET, PLAU, and TOP2A, alongside the downregulation of FABP4 and ADH1B at the protein level. Single-cell analysis revealed distinct cell-type preferences: MET in colonic epithelial cells and fibroblasts, PLAU in dendritic cells, TOP2A in proliferative epithelial and exhausted CD8+ T cells, FABP4 in vascular endothelial cells, and ADH1B in conventional dendritic cells (cDC1). CellChat analysis revealed active PLAU-mediated communication between dendritic cells and stromal populations. scTenifoldKnk-based in silico perturbation of PLAU in its native high-expression cells perturbed immune pathways, whereas MET perturbation in its high-expression cells disrupted cell adhesion and ECM-interaction pathways, suggesting their complementary roles in immune modulation and tissue architecture maintenance. CONCLUSIONS: This integrated computational study prioritized a five-gene framework potentially linking OTA-related predicted targets with CRC-associated metabolic and immune-related molecular alterations. These findings are supported by immune contexture inference, external protein-level reference data, computational structural analyses, and single-cell-based in silico perturbation results, providing a hypothesis-generating framework for subsequent experimental investigations into the potential toxicological relevance of OTA in CRC.