A de novo molecular generation method using latent vector based generative adversarial network

By Oleksii Prykhodko, Simon Johansson, Panagiotis-Christos Kotsias, Josep Arús‐Pous, Esben Jannik Bjerrum, Ola Engkvist, Hongming Chen. Journal of Cheminformatics, 2019

356 citations recorded by the source corpus.

Deep learning methods applied to drug discovery have been used to generate novel structures. In this study, we propose a new deep learning architecture, LatentGAN, which combines an autoencoder and a generative adversarial neural network for de novo molecular design. We applied the method in two scenarios: one to generate random drug-like compounds and another to generate target-biased compounds. Our results show that the method works well in both cases. Sampled compounds from the trained model can largely occupy the same chemical space as the training set and also generate a substantial fraction of novel compounds. Moreover, the drug-likeness score of compounds sampled from LatentGAN is also similar to that of the training set. Lastly, generated compounds differ from those obtained with a Recurrent Neural Network-based generative model approach, indicating that both methods can be used complementarily.

Subfields: Computational Theory and Mathematics, Materials Chemistry, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/20b248b56af9a0d1abbda300f0a275a874c7148c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Theory and Mathematics

1. Ola Engkvist
2. Tianfan Fu
3. Alán Aspuru-Guzik
4. Benjamín Sánchez-Lengeling
5. Jimeng Sun
6. Sergey I. Nikolenko
7. Yuanqi Du
8. Shengchao Liu

In Materials Chemistry

1. Ola Engkvist
2. Sergey I. Nikolenko
3. Hongming Chen
4. Tianfan Fu
5. Benjamín Sánchez-Lengeling
6. Jimeng Sun
7. Shengchao Liu
8. Yuanqi Du

In Molecular Biology

1. Ola Engkvist
2. Tianfan Fu
3. Sergey I. Nikolenko
4. Jimeng Sun
5. Alán Aspuru-Guzik
6. Hongming Chen
7. Esben Jannik Bjerrum
8. Shengchao Liu

Citations in the corpus

• Musika! Fast Infinite Waveform Music Generation (2022, neutral): “On the other hand, [24] proposes to autoencode molecules with a basic autoencoder, to then generate sequences of continuous-valued latent vectors with a GAN.”
• MolGenSurvey: A Systematic Survey in Machine Learning Models for Molecule Design (2022, methodology): “LatentGAN [159] designs a mixture model of autoencoder (AE) and GAN, where AE is used to learn a continuous latent representation of molecules and GAN encourages the generated molecules to be close to the desirable molecules.”
• Randomized SMILES strings improve the quality of molecular generative models (2019, background): “Multiple architectures have been reported in literature: the first one is Recurrent Neural Networks (RNNs) [7], but also others such as Variational AutoEncoders (VAEs) [11], Generative Adversarial Networks (GANs) [12, 13], etc.”
• Mol-CycleGAN: a generative model for molecular optimization (2019, methodology): “There are also hybrid approaches which utilize both GANs and latent vector representation in the process of compound generation [36].”
• Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models (2018, background): “, 2018), LatentGAN (Prykhodko et al., 2019), and non-neural baselines.”