Enhancing Breast Cancer Drug Development: Predicting Bioac-tivity and ADMET Properties of ERα Targeting Compounds through Machine Learning

Breast cancer is a prevalent disease affecting women globally, and discovering novel pharmacological agents with slective toxicity towards cancer cells is critical. Machine learning (ML) and molecular descriptors have become indispensable tools in drug discovery, enabling the prediction of new molecule activity prior to synthesis and experimental testing. This study aimed to enhance drug development processes by employing a hybrid approach that utilized correlation and independence analyses to identify important molecular descriptors and optimized quantitative ML models to predict bioactivity against ER$\alpha$ targets and classify compounds based on their ADMET properties. The study employed various statistical techniques, including correlation evaluation, to screen 20 descriptors related to biological activity. Based on these descriptors, the study constructed various models, including random forest, SVM, XGBoost, and BPNN, to predict biological activity and ADMET properties of compounds targeting ER$\alpha$. The study identified the six most significant molecular descriptors, along with their appropriate value ranges, to assist researchers in effectively screening small-molecule drug candidates.