Structure-Based De Novo Molecular Generator Combined with Artificial Intelligence and Docking Simulations

Abstract

Recently, molecular generation models based on deep learning have attracted significant attention in drug discovery. However, most existing molecular generation models have a serious limitation in the context of drug design wherein they do not sufficiently consider the effect of the three-dimensional (3D) structure of the target protein in the generation process. In this study, we developed a new deep learning-based molecular generator, SBMolGen, that integrates a recurrent neural network, a Monte Carlo tree search, and docking simulations. The results of an evaluation using four target proteins (two kinases and two G protein-coupled receptors) showed that the generated molecules had a better binding affinity score (docking score) than the known active compounds, and they possessed a broader chemical space distribution. SBMolGen not only generates novel binding active molecules but also presents 3D docking poses with target proteins, which will be useful in subsequent drug design.

DockStream: A Docking Wrapper to Enhance De Novo Molecular Design

“…More recently, molecular docking has been incorporated into RL generative model paradigms, offering a proposed solution that integrates structural information, steering molecular design by rewarding compounds that exhibit good docking scores and circumventing some limitations of QSAR models. [18][19][20][21] However, it is often challenging to ascertain what exactly constitutes a 'good' docking score. Docking algorithms are inherently sensitive to the three-dimensional (3D) representation of the protein and ligands [22,23].…”