Deep learning-guided discovery of IL23/IL23R macromolecular inhibitors: An integrative framework combining virtual screening and experimental validation.

Interleukin-23 (IL23) signaling, a critical driver of chronic inflammatory diseases like psoriasis, relies on the functional association of the IL23(p19) and its receptor (IL23R). Targeting this protein-protein interaction (PPI) with inhibitors presents a therapeutic advantage over biologics. Here, we report the discovery and mechanistic characterization of potent IL23/IL23R inhibitors. We employed a novel workflow starting with the Sequential Attachment-based Fragment Embedding (SAFE) deep generative model to design novel p19-targeting scaffolds, followed by virtual screening to identify commercially available compounds. Cellular screening of 31 candidates in HEK-Blue IL23 reporter cells identified Inh-31 as the optimal hit molecule. Analysis of all-atom molecular dynamics (MD) simulations (300 ns) revealed that Inh-31 binds to a deep pocket at the IL23/IL23R interface through multi-modal contacts, including crucial Pi-cation and H-bond interactions, which displace native binding hotspots. Critically, the MD data demonstrated that Inh-31 binding induces global rigidification and allosteric stabilization of the IL23R subunit (maintaining an RMSD <0.5 nm), locking the receptor complex in a non-signaling competent state to effectively halt the JAK-STAT3 cascade.