High-throughput compound screening is time and resource consuming, and considerable effort is invested into screening compound libraries, profiling, and selecting the most promising candidates for further testing. Active-learning methods assist the selection process by focusing on areas of chemical space that have the greatest chance of success while considering structural novelty. The core feature of
these algorithms is their ability to adapt the structure–activity landscapes through feedback. Instead of full-deck screening, only focused subsets of compounds are tested, and the experimental readout is used to refine molecule selection for subsequent screening cycles. Once implemented, these techniques have
the potential to reduce costs and save precious materials. Here, we provide a comprehensive overview of the various computational active-learning approaches and outline their potential for drug discovery.