DOLINA--docking based on a local induced-fit algorithm: application toward small-molecule binding to nuclear receptors.

Docking algorithms allowing for ligand and - to various extent - also protein flexibility are nowadays replacing techniques based on rigid protocols. The algorithm implemented in the Dolina software relies on pharmacophore matching for generating potential ligand poses and treats associated local induced-fit changes by combinatorial rearrangement of side-chains lining the binding site. In Dolina, ligand flexibility is not treated internally, instead a pool of low-energy conformers identified in a conformational search is screened for extended binding-pose candidates. Grouping rearranged residues in sterically independent families and side-chain conformer clustering are employed to achieve efficient use of the computational resources along with a good accuracy of the generated poses. Dolina was applied toward docking of small-molecule ligands to three different nuclear receptor ligand binding domains for which in total 18 high-resolution crystal structures were used as reference. The selected nuclear receptors feature a deeply buried ligand-binding site where local induced-fit is to be expected, particularly for receptor antagonists. For each receptor, a crystal structure with a cocrystallized small steroid ligand (template) was chosen as a target system, to which several synthetic ligands of different sizes were docked. Poses within an RMSD of 2.0 Å from the crystal reference pose were generated in 91% of the cases. In 28%, the pose with the lowest RMSD to the reference pose was ranked as the top one, and in 76% it was ranked among the top five poses. Detailed descriptions of the docking algorithm and observed results are included. Dolina is available free of charge for academic institutions.