Discovery of natural RORγt inhibitor using machine learning, virtual screening, and in vivo validation

Retinoic acid receptor-related orphan receptor gamma t (RORγt) is a crucial transcription factor regulating Th17 cells, which secrete the cytokine IL-17. RORγt inhibitors are regarded as a therapeutic modality in a wide range of autoimmunity including psoriasis. The objective of the study is to investigate novel RORγt inhibitors from natural products (NPs), combining machine learning (ML)-based virtual screening, chemotaxonomic analysis, molecular docking, and molecular dynamics simulations, and biological validation. This study employed an integrated approach combining ML-based ligand-based screening, docking study, molecular simulation, and chemotaxonomic analysis to identify RORγt inhibitors from NPs. ML ensemble models predicted potential RORγt inhibitors from an NP library; subsequent chemotaxonomic classification of top-ranked hits prioritized protoberberine alkaloids. Six protoberberine alkaloids, which are predicted to bind RORγt via docking studies, were selected for experimental validation. Among them, berberine (Ber) and coptisine (Cop) potently inhibited Th17 differentiation in vitro. Surface plasmon resonance analysis demonstrated that both Ber and Cop directly bind to RORγt, with Cop exhibiting a stronger affinity for RORγt than Ber. Moreover, Cop demonstrated therapeutic efficacy in a preclinical mouse model of psoriasis. These results validate an integrated workflow, combining ML, chemotaxonomy, and experimental testing in vitro and in vivo, for the efficient discovery of novel RORγt inhibitors.