CNS-DDI: An Integrated Graph Neural Network Framework for Predicting Central Nervous System Related Drug-Drug Interactions

The central nervous system (CNS) is one of the most complex and vital systems of the human body and is particularly interrelated with all other systems. Treatment modalities targeting the CNS as well as those targeting other systems may directly or indirectly affect the CNS. Especially in cases of polypharmacy, drug-drug interactions (DDIs) can lead to severe problems. The widespread use of drugs that have an effect on the CNS and the unpredictability of possible interactions between these drugs both complicate the treatment processes of patients and considerably increase health costs. In this study, a novel method based on Graph Convolutional Neural Networks (GCN) is proposed to predict CNS-related DDIs. The proposed approach utilizes a data fusion method by exploiting both graph structures and physical properties of drug molecules. This integrated approach enabled a more comprehensive and reliable prediction of drug interactions. The developed model achieved 98.67% accuracy and 0.994 AUC in the training process and 98.40% accuracy and 0.991 AUC in the validation process. A Graphical Interface (GUI) was designed to make the developed model easily usable by users. The integration of molecular structure and interaction network data sets a new benchmark for reliability and accuracy in DDIs prediction, addressing a critical need in modern healthcare systems. The developed methods and tools have significant potential for predicting drug interactions in the drug discovery process and in polypharmacy situations.