Navigating the frontier of drug-like chemical space with cutting-edge generative AI models

• Published in: Drug discovery today 2024-09, Vol.29 (9), p.104133, Article 104133
• Main Author: Lavecchia, Antonio
• Format: Article
• Language: English
• Subjects: artificial intelligence; chemical space exploration; de novo drug design; deep learning; drug discovery; generative models; machine learning
• ISSN: 1359-6446; 1878-5832; 1878-5832
• DOI: 10.1016/j.drudis.2024.104133

•Explore how generative AI models navigate chemical space beyond structural constraints.•Examine RNNs, VAEs, GANs, NF models, and transformers for chemical space exploration.•Discuss challenges in molecular representations, training methods, and chemical space coverage criteria.•Future focus: refine models, adopt new notations, improve benchmarks, and enhance interpretability. Deep generative models (GMs) have transformed the exploration of drug-like chemical space (CS) by generating novel molecules through complex, nontransparent processes, bypassing direct structural similarity. This review examines five key architectures for CS exploration: recurrent neural networks (RNNs), variational autoencoders (VAEs), generative adversarial networks (GANs), normalizing flows (NF), and Transformers. It discusses molecular representation choices, training strategies for focused CS exploration, evaluation criteria for CS coverage, and related challenges. Future directions include refining models, exploring new notations, improving benchmarks, and enhancing interpretability to better understand biologically relevant molecular properties.