Interaction Prediction on BACE-1 Inhibitors Data for Alzheimer Disease using Message Passing Neural Network

Abstract

The medical condition that develops as memory loss, dementia, and a general decrease in cognitive functions due to the death of brain cells over time is called Alzheimer's disease. This disease can lead to a gradual decline in cognitive functions and eventually severe memory losses that affect a person's daily life. Although the exact mechanism that causes Alzheimer's disease is not fully understood, it has been associated with certain structural changes in the brain, such as plaques and neurofibrillary bundles. This study investigates the use of geometric deep learning methods for the discovery of BACE-1 inhibitors that are promising in addressing Alzheimer's disease. Our study builds on these advancements by integrating GDL with pharmacological criteria, such as the QED criterion and Lipinski's rule, to predict BACE-1 inhibitors with enhanced accuracy and drug-like properties. Our model, which combines message-passing neural networks (MPNNs) and fully connected network (FCN) architectures, achieved a success rate of 87.7%. This performance not only surpasses that of previous studies but also ensures the practical applicability of our findings in drug discovery for Alzheimer's disease. The dual focus on prediction accuracy and drug likeness sets our work apart, providing a more comprehensive approach to identifying effective therapeutic agents.