Chapter 14 - Deep learning methodologies in drug design

Abstract

Artificial intelligence has been applied since the early 2000s in all stages of the drug design process. Recent years have witnessed a rapid growth of deep learning (DL) methodologies, which is supported by the generation of massive drug-related data and the availability of high-performance computational infrastructure. DL provides very efficient methodologies and architectures that can handle large volumes of data and are suitable for predicting drug-target interaction properties and for generating innovative molecular structures with improved properties against targets of interest. In this work, we present various DL techniques, which have been developed and applied in the field of drug design. After introducing the different approaches for creating electronic representations of molecular structures, we present the state-of-the-art DL-supported pharmaceutical discovery, including physicochemical and pharmacokinetic property prediction methods, and de novo drug design approaches based on generative DL structures and deep reinforcement learning. In the last section, we present the challenges and the limitations of the current DL technologies, to further advance DL-assisted drug discovery and design.