The development of modern machine learning technologies has highly increased the scale of computer-aided drug design in drug discovery and development. In this study we proposed a successful application of advanced deep learning technology, namely generative adversarial network (GANs), to create models for the design of druglike molecules with properties similar to known drugs. The combination of two deep neural networks, discriminator and generator, composes the GANs architectures. In the process of training, two neural networks contest with each other and were trained simultaneously. The discriminator was trained to be able to accurately discriminate between “real” and “fake” compounds derived from the generator; and the generator was trained to generate “fake” compounds to “fool” the discriminative network. The purpose of the generator is to generate compounds which will not be identified as “fake” ones by discriminator, while the purpose of the discriminator is to identify compounds coming from the generative network as “fake” ones. As for the training sets, which include about 8000 drug molecules and 10,000 screening compounds from ZINC database, four types of molecular fingerprints including FP2, FP3, FP4 and MACCS were introduced to describe the molecules with diverse structural characteristics. In addition to fingerprints, chemical SMILES descriptor was also employed to shape features of molecules. Unlike the fingerprints, this descriptor will be utilized in a recurrent neural network (RNN) to facilitate the design of druglike molecules directly. We expect the creative chemistry enabled by the advanced deep learning technique has a great application in modern drug development and discovery.

Generation of druglike molecules with generative adversarial networks (GANs)

Macromolecular Chemistry: The Second Century

[Author Benefits](https://scimeetings.acs.org/?utm_source=pubs_content_marketing&utm_medium=website&utm_campaign=0320_MCF_NPI_Launch_Spring_Homepage&ref=pubs_content_marketing)

[How to use SciMeetings](https://storage.googleapis.com/dl.morressier.com/acs/SciMeetings-ACSSpring2020-HowToGuide_M3020-Final.pdf)

[Frequently Asked Questions](https://storage.googleapis.com/dl.morressier.com/acs/SciMtng_ACSNMSp2020_FAQ_M3120_Final.pdf)

ACS Spring 2020 National Meeting & Expo

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.