Machine Learning-driven Prediction of Cervical Cancer Cell Viability After Treatment With Thymoquinone, Curcumin, and 5-Fluorouracil.

Cervical cancer remains prevalent among women globally, driven by uncontrolled cell proliferation and evasion of apoptosis. Phytochemicals like Thymoquinone (TQ) and Curcumin (CUR) are gaining attention for their safety, affordability, and biological activities. This study evaluated the cytotoxic, anti-proliferative, and apoptotic effects of TQ and CUR compared with 5-Fluorouracil (5-FU) and developed machine-learning models to predict cell viability from dose-response data. HeLa cells were exposed to increasing concentrations of TQ, CUR and 5-FU. Cell viability, morphology, and apoptosis were assessed using MTT and AO/EtBr staining. Machine-learning models: Artificial Neural Network (ANN), Support Vector Machine (SVM), Logistic Regression, Gaussian Process Regression (GPR), Random Trees (RT), and Boosted Trees (BT) were trained using experimental variables. Sensitivity analysis determined key predictors of model accuracy. Low concentrations of TQ and CUR showed minimal cytotoxicity, while higher doses reduced cell viability compared to controls. AO/EtBr staining confirmed dose-dependent apoptosis: 5-FU caused extensive late apoptosis, TQ induced both early and late apoptotic features, and CUR triggered early apoptosis. ANN demonstrated highest predictive accuracy, followed by BT and GPR. Sensitivity analysis identified IC₅₀ as the most influential parameter, with dose contributing significantly, while molecular weight had minimal impact. TQ and CUR show promising anticancer activity in cervical cancer cells. Integrating experimental assays with machine-learning models, particularly ANN, provides a framework for predicting phytochemical-based therapeutic responses and supports AI's potential in cervical cancer research.