Preselection of Compounds for Lead Identification in Virtual Screening Campaigns

Drug discovery is a multifaceted and time-intensive endeavor, traditionally involving the exhaustive experimental testing of vast libraries of chemical compounds to identify promising candidates. The advent of virtual screening has revolutionized this process, enabling the efficient identification of potential drug candidates from extensive chemical repositories. Central to the success of virtual screening is the compound filtering process, which eliminates nondrug-like compounds and enriches the library with candidates more likely to succeed. Physicochemical properties and structural descriptors play critical roles in this phase, with Lipinski’s Rule of Five (RO5) being a widely used set of filters to predict drug-likeness. Recent advancements have seen the integration of artificial intelligence (AI), machine learning (ML), and deep learning (DL) algorithms into compound filtering, enhancing accuracy and efficiency. These technologies allow for the rapid prediction of compound-target interactions and identification of structurally and functionally promising molecules. This chapter delves into the various types of filters—physicochemical, structural, pharmacophore, fragment-based, and target-based—and explores the impact of AI and ML on compound filtering, highlighting the ongoing advancements that continue to accelerate drug discovery.