Geometric Deep Learning in Drug Discovery

Abstract

Geometric Deep Learning (GDL) has significantly advanced various fields, including social networks, recommendation systems, and traffic analysis. Its application in drug discovery is a natural extension, as molecular structures have long been represented as graphs. However, unlike social networks and similar graph-based structures, molecules are inherently three-dimensional entities with unique properties that require specialized consideration. Traditional GDL methods have often overlooked these molecular-specific characteristics. My work aims to bridge this gap by integrating domain knowledge with advanced GDL techniques, creating a state-of-the-art framework to enhance the drug discovery process.