Effective virtual screening strategy toward JAK3 covalent inhibitors: combining multi‑conformational consensus calculation with covalent docking.

Accumulating studies have demonstrated that the overactivation of Janus kinase 3 (JAK3) is closely associated with various inflammatory diseases, establishing it as a potential drug target for the treatment of autoimmune and inflammatory disorders. However, the high homology among kinase structures results in poor selectivity for existing JAK3 inhibitors. The approval of the JAK3 covalent inhibitor ritlecitinib has positioned the development of covalent inhibitors as an effective strategy for enhancing JAK3 selectivity. In this study, we developed a hierarchical virtual screening cascade that includes conventional non-covalent approaches and covalent docking steps to identify novel JAK3 covalent inhibitors. First, consensus scoring-based virtual screening was performed by combining the receptor-ligand pharmacophore model with non-covalent molecular docking to pre-screen suitable non-covalently binding conformations and calculate binding energy. Subsequently, covalent molecular docking was conducted to identify molecules that can form covalent bonds with CYS909 in JAK3. This method was validated for its high accuracy while maintaining efficiency. Finally, this virtual screening strategy was employed to screen the SPECS database, resulting in the identification of several compounds with significant potential as covalent JAK3 inhibitors.