'Drug Discovery' Laboratory, Department of Pharmacy, University of Naples Federico II, I-80131 Naples, Italy.
Drug Discov Today. 2024 Sep;29(9):104133. doi: 10.1016/j.drudis.2024.104133. Epub 2024 Aug 3.
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.
深度生成模型 (GMs) 通过复杂的、不透明的过程生成新颖的分子，绕过直接的结构相似性，从而改变了对类药性化学空间 (CS) 的探索。本综述考察了 CS 探索的五个关键架构：递归神经网络 (RNN)、变分自编码器 (VAEs)、生成对抗网络 (GANs)、归一化流 (NF) 和 Transformers。它讨论了分子表示选择、针对特定 CS 探索的训练策略、CS 覆盖范围的评估标准以及相关挑战。未来的方向包括改进模型、探索新的符号、改进基准测试以及提高可解释性，以更好地理解与生物学相关的分子特性。
