Implementation of Hybrid Bat Algorithm-Ensemble on Human Oral Bioavailability Prediction of Drug Candidate

Abstract

One of significant parameters of Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) is Human Oral Bioavailability (HOB) which is crucial for determining the total of consumed drugs inside humans body circulation. Poor HOB results in undeterminable drug effects in the human body, with approximately 50% of drug candidates failing due to low oral availability. As many as 80% of drugs in the world use the oral route of entry into the body, so HOB prediction is very important to reduce side effects and the risk of toxicity brought by drugs. Unfortunately, oral bioavailability is currently predominantly measured in vivo consequently, developing in-silico methods is considered crucial. To reckon the human oral bioavailability of medication candidates, we used the Hybrid Bat Algorithm method for feature selection and the Ensemble method, i.e. Random Forest, AdaBoost, and XGBoost for the prediction model. The result showed that XGBoost as the best model in which the value of accuracy and Fl-score were 0.776, and 0.802, respectively.