Integrated in silico network toxicology, machine learning and molecular simulations to predict candidate colorectal cancer-related targets of three novel non-phthalate plasticizers
Year: 2026
Type: preprint
Abstract: Colorectal cancer (CRC) remains a major global health burden, and chronic exposure to environmental xenobiotics is increasingly implicated in colorectal tumourigenesis. Alternative non-phthalate plasticizers (DOTP, DINCH and TMCH) have been widely adopted as safer substitutes for conventional phthalates, yet their colorectal carcinogenic potential and underlying molecular mechanisms remain poorly defined. We therefore implemented an integrative computational toxicology workflow combining carcinogenicity prediction, compound-target mining, functional enrichment, weighted gene co-expression network analysis (WGCNA), large-scale machine-learning screening, multi-omics validation, immune-infiltration analysis, molecular docking and 100 ns all-atom molecular dynamics simulation. TMCH exhibited the highest predicted carcinogenic probability, and 79 genes shared between plasticizer exposure and CRC pathogenesis were identified. Among 118 machine-learning pipelines, a Random Forest plus linear discriminant analysis model using 20 of the 79 genes was selected on parsimony grounds, and SHapley Additive exPlanations prioritised five core targets: ABCG2 , CDK1 , CA7 , MET and TDP2 . Multi-cohort transcriptomics, immunohistochemistry and single-cell profiling confirmed their dysregulated expression with distinct cell-type tropism, whereas CIBERSORT and ESTIMATE analyses linked these genes to specific immune-cell subsets and to the global immune-stromal microenvironment. All five genes achieved strong diagnostic performance but showed divergent prognostic associations. Docking predicted favourable binding of DOTP-ABCG2 and DINCH-CDK1, whereas TMCH bound weakly to all five proteins despite its high predicted risk; molecular dynamics simulations supported the structural stability of the DOTP-ABCG2 complex. Collectively, this in silico study provides testable hypotheses rather than causal proof and nominates five candidate targets for experimental toxicological validation.
Source: Research Square
Authors: Wenzhu Lou, Wei Ren, Bangsheng Chen
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7225432489