Prediction of Multi Class Drugs: A Perspective for Designing Drug with Many Uses

Abstract

The drug-like molecule which could treat multiple diseases is commercially more viable and can act on multiple biological pathways. Such drug candidates can also be more important in the treatment of complex diseases like cancer. Traditional methods are not focused on the development of such drugs, but computational method can be developed to predict multiple disease potential of drug-like molecules. Computational methods have been extremely successful in drug discovery through prediction of drug potential of the drug-like molecules such as toxicity, physiological effects, binding energy and binding pose with the receptor. Computational methods to predict multiple disease potential of the drug-like molecules are not worked out so far in spite of the high importance of such drugs and it can also expedite the drug repurposing. Hence, information of approved drugs used for the treatment of single and multiple diseases was included to develop the machine learning-based model for the prediction of multiple disease potential of the drug-like molecules. Molecular descriptors were used as the features and optimally selected for support vector machine-based prediction models. The fairly high accuracy of developed method justifies the importance of selected method and approach. The developed method is expected to expedite the drug discovery process through the prediction of multi-drug potential of drug-like molecules.