The use of DNA-encoded libraries (DELs) in drug discovery efforts has gained significant popularity in recent years as an alternative to traditional high-throughput screening (HTS) methods. These libraries consist of billions of molecules which are linked with unique DNA identifiers. DEL affinity screens assess potential hits by counting the occurrence of each DNA sequence after a bind-wash-elute cycle against the desired target. However, the data obtained from these experiments is inherently noisy, varying widely based on library quality and target conditions, and therefore finding true hits can be challenging.<br/>In this presentation, we describe a DEL data analysis platform that utilizes unsupervised and supervised machine learning models to procedurally identify diverse chemical series for follow-up efforts and derive structure activity relationship insights from DEL screening data. These tools collectively provide reliable hit-calling considering false positives and negatives which is a limitation in conventional visual mining techniques. Furthermore, the analysis pipeline takes advantage of existing open-source software libraries and has been made accessible to chemists through our existing informatics platform. Through this, we have realized an unbiased method of processing DEL data which can be applied uniformly to assays with varying conditions, targets, and libraries.

A DEL data analysis platform for the identification of diverse chemical series in drug discovery

ACS Fall 2021

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