The Drug Design for Diabetes Mellitus type II using Rotation Forest Ensemble Classifier

Abstract

Abstract Dipeptidyl peptidase-IV (DPP-IV) inhibitor is one of the drug targets for the treatment of diabetes. Some classes of those drugs have dangerous side effects so is critical to develop safer drugs. By using rotation forest methods and in silico, it will be more efficient than conventional methods that require a lot more costs and are more time-consuming. One of in silico methods used in drug design is ligand-based virtual screening (LBVS). The interlocking structure capabilities are identified by the LBVS Process. The fingerprint is one of the structural interpretations. Molecular fingerprints are used as a criterion for LBVS in computational drug discovery. A circular fingerprint is found to improve LBVS performance. In this paper, we used the representation of ECFP and FCFP as a method to extract features, after which we used a Rotation Forest classifier to predict active and inactive compounds. The experiment result shows Rotation Forest has good prediction based on the different circular fingerprint and can successfully better classify with results of MCC being 85% and accuracy 92%.