Application of Stacking and Blending Ensemble Strategies in Breast Cancer Drug Candidates

Abstract

In order to select drugs that are effective against breast cancer and can be absorbed safely by humans, the composition and molecular structure of the drugs need to be analyzed. A regression prediction model based on the Blending strategy is proposed for the bioactivity of ER$\alpha$, the target of breast cancer treatment, and a classification prediction model based on the Stacking strategy is proposed for the ADMET(Absorption, Distribution, Metabolism, Excretion, Toxicity) properties to optimize the bioactivity and ADMET properties of ER$\alpha$ antagonists. The experimental results showed that 23 molecular structures with significant effects on bioactivity were selected by regression models, and the ensemble method based on the Blending strategy could better fit whether the molecular structure could inhibit ER$\alpha$ activity than other comparative models. The ensemble method based on the Stacking strategy was able to predict more accurately whether the drug could be a breast cancer drug candidate than other comparative models.