Covalentizer – An automatic pipeline for the generation of irreversible derivatives, discovers a nanomolar covalent ERK2 inhibitor

Design of covalent inhibitors is a challenge of increasing importance in drug discovery. Covalent inhibitors can display higher affinity, prolonged residence time, and a potential for greater specificity than their non-covalent counterparts. Efficiently designing covalent inhibitors, though, is still a great challenge. Experimental high throughput screening, is unheard of for covalent binders. Fragment screening and virtual screening are gaining popularity, but both suffer from a common drawback - the typically low affinity of initial hits. This drawback can lead to very lengthy and expensive optimization campaigns. In this work we present Covalentizer, a computational pipeline for creating irreversible inhibitors, based on X-ray structures of known reversible binders. We curated all of the protein structures within the PDB of a ligand which binds near a free cysteine and found 8,386 such structures. For each of these non-covalent ligands, we created a custom-made and focused library of covalent analogs. We used DOCKovalent, a covalent docking method, to dock these tailored covalent libraries and calculated their RMSD to the crystalographic pose, to find those which can bind covalently to the nearby cysteine while keeping the main interactions of the original molecule. For 1,553 structures we found at least one covalent derivative which docks within 1.5Å RMSD to the reversible crystallographic ligand. We tested selected predictions from this wide-scale collection and among them, discovered two new covalent inhibitors for ERK2, one of which with an IC₅₀ of 3.4nM. This suggests that many of the predictions represent novel and potent covalent inhibitors, for a very wide range of protein targets.