Machine Learning-Driven QSAR Modeling of Anticancer Activity from a Rationally Designed Synthetic Flavone Library

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Flavones are recognized as privileged scaffolds in drug discovery, showing significant promise as anticancer agents.
• Optimizing lead compounds for anticancer drug development requires efficient methodologies.
• Quantitative structure-activity relationship (QSAR) models can accelerate the identification of potent drug candidates.

Purpose of the Study:

• To develop and validate a machine learning-driven QSAR model for anticancer flavone derivatives.
• To design and synthesize novel flavone analogs with enhanced cytotoxicity against cancer cell lines.
• To identify key molecular descriptors that influence the anticancer activity of flavones.

Main Methods:

• Design and synthesis of 89 flavone analogs using pharmacophore modeling.
• Biological evaluation of synthesized compounds against breast (MCF-7) and liver (HepG2) cancer cell lines, and normal Vero cells.
• Development and comparison of machine learning models (Random Forest, Extreme Gradient Boosting, Artificial Neural Network) for QSAR analysis.
• Validation of the best-performing model using test compounds and SHapley Additive exPlanations (SHAP) for descriptor analysis.

Main Results:

• Promising flavone candidates exhibited enhanced cytotoxicity against MCF-7 and HepG2 cancer cells with low toxicity to normal cells.
• The Random Forest (RF) model demonstrated superior performance, achieving R² values of 0.820 (MCF-7) and 0.835 (HepG2).
• Cross-validation (R²cv) and test set validation confirmed the robustness of the QSAR model, with RMSEtest values of 0.573 (MCF-7) and 0.563 (HepG2).
• SHAP analysis identified critical molecular descriptors influencing anticancer activity, aiding in rational drug design.

Conclusions:

• A robust machine learning-driven QSAR model was successfully developed for anticancer flavone derivatives.
• The model facilitates the rational design of selective and potent anticancer agents.
• This approach accelerates the optimization of lead compounds in anticancer drug discovery.