Mechanism-Based Multitarget Modeling for Pathway-Level Prediction of PI3K/Akt Signaling Perturbation Induced by Liquid Crystal Monomers

Liquid crystal monomers (LCMs) are emerging contaminants whose system-level toxicity mechanisms remain poorly understood. Here, we developed a pathway-centric multitarget framework to characterize coordinated toxicological perturbations at the signaling network level. KEGG enrichment identified the PI3K/Akt pathway as a key mechanistic axis, and a minimal set of 19 proteins covering upstream receptors, central kinases, and downstream effectors was constructed. A multitask deep learning model trained on ChEMBL IC 50 data (53 694 molecules; 62 440 data points) achieved strong performance (accuracy >0.85, up to 0.94) and was applied to 1412 LCMs. EGFR/JAK1 showed up to 20.18% predicted active inhibitors, followed by PIK3CA (18.06%), with some LCMs exhibiting docking energies comparable to known inhibitors. Structural analysis identified fluorinated aromatic rings, cyclohexyl and bicyclic aliphatic scaffolds, and oxygen-containing groups as key contributors to pathway perturbation. Integration of pathway-level inhibition profiles with simulated Gene Ontology enrichment linked multinode interference to cell cycle arrest, apoptosis, immunosuppression, and impaired cell migration. Transcriptomic analysis in human lung epithelial (A549) cells further confirmed the predicted disruption of the PI3K/Akt signaling pathway. This study provides a transferable framework for mechanism-informed toxicity assessment and chemical prioritization under limited experimental data.