Prediction of new bioactive molecules of chemical compound using boosting ensemble methods

Abstract

Virtual screening (VS) methods can be categorized into structure-based virtual screening (SBVS) that involves knowledge about the target’s 3D structure and ligand-based virtual screening (LBVS) approaches that utilize information from at least one identified ligand. However, the activity prediction of new bioactive molecules in highly diverse data set is still less accurate and the result is not comprehensive enough since only one approach is applied at one time. This paper aims to recommend the boosting ensemble method, MultiBoost, into LBVS using the well-known chemoinformatics database, the MDL Drug Data Report (MDDR). The experimental results were compared with Support Vector Machines (SVM). The final outcomes showed that MultiBoost ensemble classifiers had improved the effectiveness of the prediction of new bioactive molecules in high diverse data.