Molecules Absorption Prediction Using Support Vector, Adaboost, Random Forest and Decision Tree Classification

Abstract

Classification is supervised machine learning applicable to predict chemicals based on their properties. The chemical properties are derived from its structural and functional groups. Many molecular descriptors have been developed, one of which, was pharmacophore. Pharmacophore is a quantitative measure of molecules in their application as a pharmaceutical ingredient. The training datasets were 59 molecules categorized on their adsorption properties. The classification was carried out to divide the training set into their adsorption class using their pharmacophores. The prediction of enolic curcumin and its degradation product was used to verify the trueness of classification methods based on their pharmacophores. Curcumin and its degradation product were used because there were many studies carried out about curcumin and its pharmaceutical effect.