We describe a novel deep learning neural network method and its application to impute compound bioactivities and properties [1]. Unlike conventional machine learning approaches, this method can train directly on sparse, noisy bioactivity data, typical of those available in drug discovery. In combination with molecular descriptors, this enables it to learn directly from correlations between activities measured in different assays as well as structure-activity relationships. Furthermore, the model provides a robust estimate of the confidence in each prediction, enabling attention to be focused on only the most accurate results. We present a case study that demonstrates that the neural network method outperforms traditional quantitative structure-activity relationship models and discuss how these results can be used to fill in missing data, predict compound activity profiles and identify new active compounds.<br/><br/>

Practical applications of deep learning to imputation of drug discovery data

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.