Malkeet Singh Bahia, Yifan Jiang, Arthur Gonse, Robert A. Batey, Michał Koziarski

Digital Discovery

DOI: 10.1039/d6dd00117c

Abstract

Over the past several decades, DNA-encoded libraries (DELs) have become an established and efficient approach for the synthesis and screening of large molecular libraries to facilitate drug discovery. Although DELs generate substantially larger amounts of protein–ligand binding data than traditional screening approaches, the quality of the resulting molecular enrichment counts is often poor. Consequently, a careful treatment of enrichment counts is often required to use DELs for further analysis. This review examines different sources of noise influencing DEL screening data quality, and strategies for de-noising enrichment counts, including ML frameworks. Furthermore, it summarizes successful DEL-ML case studies for hit identification and benchmarks for different ML methods.