Compound screening is a vital and impactful stage of non-target based early drug discovery. Achieving a high hit rate demands intelligent selection of the input libraries. If the chosen scaffolds have high potency, low toxicity, and high diversity, the drug discovery process has a higher chance of success. Virtual screening has proven to increase the efficiency of the early stage via predicting the activity of compounds in-silico. With the introduction of deep learning models to virtual screening, the accuracy of the models have often improved yet the interpretability of the predictions have decreased. This fact presents a dilemma in using deep learning models for compound selection. The higher accuracy increases the reliability of the selection, yet the lack of interpretability hinders the experts’ decision making. Therefore, being able to rely on the deep learning model’s prediction and interpret the output are crucial in employing the trained model in compound selection. In this work, a practical guideline is presented for using network’s predictions in order to select libraries for screening. This guideline manages the output of the model, the confidence of the prediction, and interpretation of the results. A pragmatic training pipeline is presented alongside the selection guideline which includes transfer learning and external validation to demonstrate how training can be improved to attain better selections. Overall, the presented approach results in the selection of diverse compounds with high probability of activity and low chance of toxicity, enabling the use of deep learning models in drug repurposing or early stages of non-target based drug discovery.

Application of deep learning in purchasing compound libraries for non-target based 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.