Accurate physics-based flexible docking of macrocyclic ligands

Macrocycles are an increasingly relevant class of therapeutics, but their size and highly constrained conformational space poses unique challenges to computational modeling. Existing modeling protocols struggle to accurately sample macrocycle bioactive conformations and predict their binding mode to receptors of interest, hindering structure-based drug discovery campaigns. To address these limitations, we developed a new flexible macrocycle docking protocol based on existing Schrödinger technologies for macrocycle sampling and small molecule docking. To benchmark its accuracy, we collected a dataset of 240 diverse receptor-macrocycle systems, expanding and improving over previously published datasets. Our results show that our method improves significantly over existing docking tools, recapitulating the receptor-bound structure in the top 2 ranked poses for 82% of cases, including several clinically relevant compounds. Finally, we discuss the limitations of our method and the remaining challenges of macrocycle modeling, such as induced-fit effects, strain, and the accurate placement of chemical decorations around the macrocycle core.