Diverse Computational Strategies Enable the Discovery of p38α−MK2 Molecular Glues

Abstract Mitogen-activated protein kinase-activated protein kinase 2 (MK2) phosphorylates AU-rich element-binding proteins (AUBPs) that modulate the stability of cytokine mRNA. An emerging class of inhibitors creates a ternary complex between upstream kinase p38α and MK2. These molecular glues demonstrate sustained efficacy relative to orthosteric inhibitors of either p38α or MK2; however, their mechanism of action presents challenges to computationally driven design. Herein we describe in silico hit identification strategies leveraging free energy perturbation coupled with machine learning to triage and prioritize ligands from diverse chemical sources and identify multiple structurally diverse, potent, and highly selective p38α−MK2 molecular glues. Lead compounds 14 and 28 demonstrated an attractive PK profile with sufficient coverage of human whole blood EC50 and displayed pronounced reduction of TNFα levels after PO dosing in lipopolysaccharide mouse models, illustrating the value of our in silico hit identification strategies and optimization of non-degrading molecular glues.