Deep learning has recently made advances in a variety of important tasks in computational drug discovery. With the advent of deep generative models, these methods can now not only be used for classification and regression, but also for generating new realistic data. Previously, our lab developed convolutional neural networks (CNN) that effectively discriminate between active molecules and decoys for a target protein of interest, as well as between correct and incorrect binding poses, by using a grid-based three-dimensional atomic density representation of protein-ligand complexes as input.<br/> We now extend from protein-ligand scoring to <i>de novo </i>drug design by training models that output this atomic density representation using only the receptor structure as input. Critically, a generative model can be designed such that it can be sampled conditionally. For example, a ligand structure can be sampled conditioned on the desired binding site. We have developed deep generative models including variational autoencoders (VAE) and conditional generative adversarial networks (GAN) to learn a latent space that allows us to direct sample the distribution of active ligand densities. Furthermore, since the distribution learned by a generative model may not map to all feasible molecular structures or may map to mostly infeasible structures, there remain many challenges. We will show our novel molecular validity methods which evaluate how well the generated grid densities match to the expected densities of a realistic molecule and discuss them in our poster.

Deep generative models for computational drug discovery

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.