The Selectivity Implications of Docking Libraries with Greater and Lesser Similarities to Bio-like Molecules

As make-on-demand libraries have expanded into the billions, their similarity to "bio-like" molecules has rapidly diminished. Nevertheless, docking these ultralarge libraries has found actives with high hit rates and affinities. Plausibly, the divergence from bio-like may improve hit selectivity; conversely, if hit rates on target are divorced from bio-like similarity, off-target selectivity may be as well. Here, we test whether docking actives from ultralarge libraries are more selective than those from more bio-like "in-stock" libraries in an in-depth study against a single target, the 5-HT_2A receptor. While chemoinformatics predicts more off-targets for in-stock actives, when tested experimentally against 318 GPCRs, both 16 agonists from the ultralarge and 20 ligands from the in-stock libraries had similar selectivities. The ultralarge library agonists are more subtype-selective within the 5-HT₂ family; overall, these results suggest that off-target selectivity, like affinity and hit rates for on targets, is divorced from library similarity to bio-like molecules.