Discovery of Covalent Ligands with AlphaFold3

Area of Science:

• Biochemistry
• Computational chemistry
• Drug discovery

Background:

• Covalent inhibitors are crucial research and therapeutic tools.
• Limited computational methods hinder the discovery of covalent inhibitors.
• AlphaFold3 (AF3) shows promise in ligand pose prediction but requires validation for virtual screening.

Purpose of the Study:

• Assess the applicability of AlphaFold3 (AF3) for virtual screening of covalent inhibitors.
• Evaluate AF3's performance against state-of-the-art covalent docking tools.
• Discover novel covalent small molecules targeting kinases using AF3.

Main Methods:

• Utilized AF3 co-folding predictions and confidence metrics for ranking covalent binders.
• Performed virtual screening against the Bruton's tyrosine kinase (BTK).
• Validated binding mode accuracy using co-crystallography.

Main Results:

• AF3 predictions effectively ranked true covalent binders over decoys, outperforming existing covalent docking tools.
• A novel, chemically distinct covalent small molecule inhibitor of BTK was discovered.
• In vitro and cellular assays confirmed potent inhibition by the novel molecule.
• Co-crystallography confirmed sub-angstrom accuracy of AF3-predicted binding modes.

Conclusions:

• AlphaFold3 (AF3) is a practical tool for discovering novel covalent chemical matter for kinase targets.
• AF3-driven virtual screening accelerates the identification of potent and selective covalent inhibitors.
• This AI approach significantly advances covalent inhibitor drug discovery for kinases.