Drug Likeness Prediction Using Structure Based Molecular Descriptors and Support Vector Machines

Abstract

Abstract In the drug development process, the selection of potential drug candidate molecules, by virtual screening, often depend on computation intensive methods. The processing objectives mainly focus on the classification, structure analysis and segregation of molecules using numerical data. In this direction, a screening strategy is designed which makes use of the numerical data derived from the geometry of the molecular structures and with a data mining algorithm called Support Vector Machine (SVM).For implementing the algorithm, Radial basis function and Neural kernels are used to predict the possibility of drug like characteristics in the test molecules. The results of the prediction experiment have been compared with the predictions made by Drug Mint, a web server tool