Graph machine learning in drug discovery

Over the last decade, we have witnessed deep learning's (DL) revolutionizing countless fields ranging from image processing to natural language processing. While most of the earlier applications of DL were well-structured Euclidean domains, recently there has been an increasing desire to extend DL approaches on nonEuclidean data structures such as graphs. This has led to the introduction of Graph Machine Learning (GML), a class of machine learning algorithms that operate on graphs. Graphs are ubiquitous in nature and provide a robust mathematical abstraction for numerous real-life entities including biomolecules and drugs. As a result, the technical advances in the field of GML have attracted keen interest from pharmaceutical companies to explore their usefulness in Computer-Aided Drug Discovery (CADD) tasks. In this article, we systematically review the applications of GML in CADD with a focus on proteins and small drug molecules.