Docking large libraries of molecules against unrefined AlphaFold2 (AF2) models of the σ2 and serotonin 2A (5-HT2A) receptors produced hit rates and affinities as high as those obtained by docking against experimental structures. These results were achieved despite differences in orthosteric residue conformations between the AF2 models and the experimental structures. A cryo–electron microscopy structure of one potent 5-HT2A ligand identified from AF2 docking showed residue accommodations similar to the AF2 prediction. AF2 models may sample low-energy conformations that differ from experimental structures but remain useful for ligand discovery, extending the reach of structure-based drug design.
The serotonin transporter (SERT) adopts three conformations, and most antidepressants target its outward-open state. Ibogaine, which targets the inward-open state, has an unusual antidepressant profile but is cardiotoxic. Computational docking of over 200 million small molecules against the ibogaine-stabilized inward-open SERT identified 36 top compounds; 13 inhibited SERT with potencies from 29 to 5000 nM. Optimization yielded two inhibitors with Ki values as low as 3 nM that stabilized an outward-closed state and showed little off-target activity. A cryo-EM structure confirmed the predicted binding geometry. In mice, both compounds showed anxiolytic and antidepressant activity with potencies up to 200 times greater than fluoxetine.