Based on the training strategies, deep molecular generative models can be classified into two categories: reinforcement learning (RL)-based methods, which generate molecules with desired properties; unsupervised (UL)-based or self-supervised (SSL)-based methods, which aim to generate valid, novel, and diverse molecules; supervised (SL)-based methods generating molecular three-dimensional conformations from molecular graphs.
- [Front. Pharmacol. 2020] [MOSES] Daniil, Polykovskiy, et al. Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models
- [NeurIPS 2018] [GCPN] You, Jiaxuan, et al. Graph convolutional policy network for goal-directed molecular graph generation [code]
- [Sci. Adv. 2018] [ReLeaSE] Popova, Mariya, et al. Deep reinforcement learning for de novo drug design [code]
SL-based generator - molecular conformation
- [ICLR 2022 (Oral)] [GeoDiff] Xu, Minkai, et al. Geodiff: A geometric diffusion model for molecular conformation generation [code]
- [NeurIPS 2022] [torsional diffusion] Jing, Bowen, et al. Torsional diffusion for molecular conformer generation [code]
- [TMLR 2022] [DMCG] Zhu, Jinhua, et al. Direct molecular conformation generation [code]
- [NeurIPS 2021] [GeoMol] Ganea, Octavian, et al. Geomol: Torsional geometric generation of molecular 3d conformer ensembles [code]
- [ICML 2021] [ConfGF] Shi, Chence, et al. Learning gradient fields for molecular conformation generation [code]
- [ICML 2021] [ConfVAE] Xu, Minkai, et al. An end-to-end framework for molecular conformation generation via bilevel programming [code]
- [ICLR 2021] [CGCF] Xu, Minkai, et al. Learning neural generative dynamics for molecular conformation generation [code]
- [NeurIPS 2020] [TorsionNet] Gogineni, Tarun, et al. Torsionnet: A reinforcement learning approach to sequential conformer search [code]
- [ICML 2020] [GraphDG] Simm, Gregor NC, and José Miguel Hernández-Lobato. A generative model for molecular distance geometry [code]
- [Sci. Rep. 2019] [CVGAE] Mansimov, Elman, et al. Molecular geometry prediction using a deep generative graph neural network [code]
UL-based & SSL-based generator - molecular graph
- [ICLR 2022 (Oral)] [DEG] Guo, Minghao, et al. Data-efficient graph grammar learning for molecular generation [code]
- [ICML 2021] [GraphDF] Luo, Youzhi, et al. Graphdf: A discrete flow model for molecular graph generation [code]
- [ICML 2020] [RationaleRL] Jin, Wengong, et al. Multi-objective molecule generation using interpretable substructures [code]
- [ICLR 2020] [GraphAF] Shi, Chence, et al. Graphaf: A flow-based autoregressive model for molecular graph generation [code]
- [ICML 2020] [HierVAE] Jin, Wengong, et al. Hierarchical generation of molecular graphs using structural motifs [code]
- [arXiv 2019] [GraphNVP] Madhawa, Kaushalya, et al. Graphnvp: An invertible flow model for generating molecular graphs [code]
- [NeurIPS 2018] [CGVAE] Liu, Qi, et al. Constrained graph variational autoencoders for molecule design [code]
- [NeurIPS 2018] Ma, Tengfei, et al. Constrained generation of semantically valid graphs via regularizing variational autoencoders 😢 No official codes are available.
- [ICML 2018] [JT-VAE] Jin, Wengong, et al. Junction tree variational autoencoder for molecular graph generation [code]
- [ICML 2018] [MolGAN] De Cao, Nicola, and Thomas Kipf. MolGAN: An implicit generative model for small molecular graphs [code]
UL-based & SSL-based generator - SMILES string
- [Chem. Sci. 2021] [STONED] Nigam, AkshatKumar, et al. Beyond generative models: superfast traversal, optimization, novelty, exploration and discovery (STONED) algorithm for molecules using SELFIES [code]
- [arXiv 2018] [ORGAN] Guimaraes, Gabriel Lima, et al. Objective-reinforced generative adversarial networks (organ) for sequence generation models [code]
- [J Chem Inf Model 2018] [BIMODAL] Grisoni, Francesca, et al. Bidirectional molecule generation with recurrent neural networks [code]
- [ACS Cent. Sci. 2018] [VSeq2Seq] Gómez-Bombarelli, Rafael, et al. Automatic chemical design using a data-driven continuous representation of molecules [unofficial code] 😢 No official codes are available.
- [ACS Cent. Sci. 2018] Segler, Marwin HS, et al. Generating focused molecule libraries for drug discovery with recurrent neural networks [unofficial code] 😢 No official codes are available.
- [ICML/PMLR 2017] [GVAE] Kusner, Matt J., et al. Grammar variational autoencoder [code]
While both molecular generation and optimization involve creating new molecules, generation is focused on creating entirely new molecules from scratch, while optimization is focused on improving the properties of existing molecules.
- [ICLR 2019] Jin, Wengong, et al. Learning multimodal graph-to-graph translation for molecular optimization [code]
- [Sci. Rep. 2019] [MolDQN] Zhou, Zhenpeng, et al. Optimization of molecules via deep reinforcement learning [code]