Macrocycles are small molecules that contain at least one large, organic ring. These compounds are a growing class of therapeutics that now represents 4% of total FDA approvals. Computational modeling of macrocycles is challenging due to their size and correlated internal degrees of freedom, both of which make conformational sampling difficult. Current modeling tools struggle to accurately recapitulate bioactive conformations and predict their binding mode inside protein receptors. To address this, we assembled the largest macrocycle docking data set to date and developed a new flexible macrocycle docking technology, termed MacroDock. Our data set comprises 240 high-quality macrocycle-receptor complexes with diverse ligand chemistry, size and receptor identity. We demonstrate that MacroDock is an accurate docking protocol and produces a sub-2.0 Å top 2 ranked pose for 80.1% of cases. Additionally, we highlight successful docking of clinically relevant compounds and discuss remaining challenges for macrocycle docking, such as ligand-induced strain. MacroDock represents an accurate and efficient solution for macrocycle docking, and will have important implications for prospective structure-based drug discovery.
Robson‐Tull et al. (Sun,) studied this question.