Synergistic acceleration of machine learning and molecular docking for prostate-specific antigen ligand design

Prostate-specific antigen (PSA) serves as a critical biomarker for the early detection and continuous monitoring of prostate cancer. However, commercial PSA detection methods primarily rely on antigen-antibody interactions, leading to issues such as high costs, stringent storage requirements, and potential cross-reactivity due to PSA variant sequence homology. This study is dedicated to the precise design and synthesis of molecular entities tailored for binding with PSA. By employing a million-level virtual screening to obtain potential PSA compounds and effectively guiding the synthesis using machine learning methods, the resulting lead compounds exhibit significantly improved binding affinity compared to those developed before by researchers using high-throughput screening for PSA, substantially reducing screening and development costs. Unlike antibody detection, the design of these small molecules offers promising avenues for advancing prostate cancer diagnostics. Furthermore, this study establishes a systematic framework for the rapid development of customized ligands that precisely target specific protein entities. By employing a million-level virtual screening to obtain potential PSA compounds and effectively guiding the synthesis using machine learning methods, we obtained lead compounds that exhibited significantly improved binding affinity for PSA.