Deep Generative Models in De Novo Drug Molecule Generation

Abstract

The discovery of new drugs has important implications for human health. Traditional methods for drug discovery rely on experiments to optimize the structure of lead molecules, which are timeconsuming and high-cost. Recently, artificial intelligence has exhibited promising and efficient performance for drug-like molecule generation. In particular, deep generative models achieve great success in de novo generation of drug-like molecules with desired properties, showing massive potential for novel drug discovery. In this study, we review the recent progress of molecule generation using deep generative models, mainly focusing on molecule representations, public databases, data processing tools, and advanced artificial intelligence based molecule generation frameworks. In particular, we present a comprehensive comparison of state-of-the-art deep generative models for molecule generation and a summary of commonly used molecular design strategies. We identify research gaps and challenges of molecule generation such as the need for better databases, missing 3D information in molecular representation, and the lack of high-precision evaluation metrics. We suggest future directions for molecular generation and drug discovery.

Artificial Intelligence-Assisted Optimization of Antipigmentation Tyrosinase Inhibitors: De Novo Molecular Generation Based on a Low Activity Lead Compound

“…10 However, previous RL models based on "atom and bond" growth often produce molecules with unrealistic structures that lack synthesizability and reasonable drug-like properties and therefore are often unable to verify their biological activity. 11 To overcome these limitations, researchers have put forward fragment-based algorithms. 12−14 Following the methodology proposed in ref 14, we have developed a fragment-by-fragment RL forward synthesis algorithm for generating molecules based on the lead compound and reaction templates.…”

“…Due to its remarkable capacity for exploration and learning, RL has gained popularity in the field of multiobjective optimization for drug molecular design . However, previous RL models based on “atom and bond” growth often produce molecules with unrealistic structures that lack synthesizability and reasonable drug-like properties and therefore are often unable to verify their biological activity . To overcome these limitations, researchers have put forward fragment-based algorithms. − …”

CLigopt: Controllable Ligand Design Through Target-Specific Optimisation

“…Attributed to the success of syntactic pattern recognition in language modelling tasks, early molecular generative models are designed on the basis of SMILES, a line notation that describes chemical structures using ASCII strings. However, SMILES representation is limited by its intrinsic instability in describing molecular structural information, as structurally similar molecules can result in dissimilar SMILES strings [4].…”

GenerRNA: A generative pre-trained language model forde novoRNA design

“…For instance, ProGen, a large-scale protein language model, generates protein sequences with predictable functions across diverse protein families, also aiding in the exploration of protein space [7]. Similarly, in the realm of small molecule design, there has been a proliferation of published generative language models utilizing string-based molecule representations, garnering considerable attention [8]. As for RNA research, while pre-trained language models have demonstrated efficacy in RNA function and structure prediction [9], the generation of RNA sequences remains largely unexplored.…”