Shared interaction pathways of ligands targeting the ligand-binding pocket of nuclear receptors

Nuclear receptors (NRs) have been exploited as one of the most important target families for clinical therapies. However, the closed structure of ligand-binding domain (LBD), the main region for NR drug design, usually hinders researchers in understanding the kinetic process of drug-target interaction. In this study, by combining extensive enhanced sampling simulations and machine learning (ML), we comprehensively explore the kinetic features (dissociation pathways) of ligands targeting 21 clinically engaged NRs. Detailed analysis shows that the conservation of protein structures is the primary factor influencing the pathway selection of an NR ligand. Furthermore, a machine-learning model (MIEC-ResMLP) is constructed to predict the pathway distribution of the NR ligands, which shows reliable prediction capability ( r p > 0.7) in various cases. This study provides the most comprehensive up-to-date analysis for interaction pathways of NR ligands, and the results may help NR drug development through better kinetic properties.