AlphaFold2 structures guide prospective ligand discovery.

Area of Science:

• Computational biology
• Structural biology
• Drug discovery

Background:

• AlphaFold2 (AF2) models have shown broad utility but variable success in retrospective ligand recognition studies.
• Structure-based drug design relies on accurate protein target structures for effective ligand docking.

Purpose of the Study:

• To prospectively evaluate the utility of unrefined AlphaFold2 models for ligand recognition against σ2 and serotonin 2A (5-HT2A) receptors.
• To compare the performance of AF2 models with experimental structures in docking large compound libraries.

Main Methods:

• Prospective molecular docking of large compound libraries against unrefined AF2 models of σ2 and 5-HT2A receptors.
• Comparison of hit rates and ligand affinities obtained from docking against AF2 models versus experimental structures.
• Determination of cryo-electron microscopy structure for a potent 5-HT2A ligand identified via AF2 docking.

Main Results:

• High and comparable hit rates and affinities were achieved using both experimental and AF2-derived structures.
• Successful ligand docking against AF2 models occurred even with differing orthosteric residue conformations compared to experimental structures.
• Cryo-EM analysis of a top ligand revealed residue accommodations consistent with AF2 predictions.

Conclusions:

• Unrefined AlphaFold2 models can accurately predict ligand binding, even with conformational differences from experimental structures.
• AF2 models represent low-energy, relevant conformations suitable for structure-based drug design and ligand discovery.
• These findings expand the applicability of AF2 models in drug design, complementing traditional structure-based methods.