Target-Disease-Drug Association Network- Guided Classification and Drug Repurposing of Neurological Disorders

Abstract

Accurate therapeutic intervention against many Neurological disorders is still not known. Only symptomatic treatments are being used for the cure of such devasting disorders. Therefore, it is crucial to probe the target associations with the drug followed by the subsequent disease associations which could aid in more accurate and effective treatment against neurological disorders. Additionally, there is a major overlap of targets in neurological and neurodegenerative disorders. In this study, a database for known protein-targets and FDA-approved drugs for 10 neurodegenerative disorders and 9 neurological disorders is developed from publicly available resources. The database contains 236 unique protein-targets with Protein-Protein Interactions (PPIs) ranging from 3 to 71, and 964 FDA-approved drugs against selected target-proteins for the 19 neuronal disorders. Network pharmacology approach was used to investigate the targets association and overlap in neurological and neurodegenerative disorders. Three networks i.e.,

Target-Disease, Disease-Drug and Target-Disease-Drug Networks, were built between protein- targets, FDA-approved drugs, and neuronal disorders, with datasets categorized into neurological

and neurodegenerative disorders. Furthermore, five machine learning models were trained on the networks, with Decision Tree, Random Forest, and Gradient Boosting Classifiers emerging as optimal models for predicting disease association of protein-targets and drugs. The results provide a comprehensive view of drugs and protein-targets’ association with specific neurological and neurodegenerative disorders, as well as target overlap among multiple neuronal disorders. Finally, a multi-variate Artificial Neural Network (ANN) to predict drug-target interactions linked to specific diseases has been developed. The model was trained using a multi-variate output configuration, enabling predictions for both target protein descriptors with 53% accuracy and disease class with 82% accuracy, for a given drug. This study contributes to database development and Network classification for FDA-approved drugs and protein targets associated with neurological and neurodegenerative disorders including the multi-variate model development, offering potential avenues for developing new therapeutics and personalized treatment strategies.