The tumor sections were also analyzed immunohistochemically using anti-vWF antibody

ues, such as Asp233, for chemotype diversity, and for compounds that were unburdened by known problems of the DOCK3.6 protocol and scoring, pricipally incorrect ionization and tautomerization states of the docked molecules, and occasionally high-internal energy conformations, as previously described33. Binding was detected for 14 of these 15 with affinities ranging from 8 to 982 mM . Crystal structures of Gateless in complex with six of the new dockingpredicted ligands were determined with resolutions ranging from 1.30 to 1.86 A. The structures of two ligand complexes, those of compounds 10 and 17, superposed to within 0.5 A of the docking prediction, three structures did to within 1.4 A of the docking prediction, and for one ligand the docking pose was over 3 A away from the crystallographic result. The crystallographic orientations of compounds 14 and 24 differed mainly by a translational Affects of Water on Ligand Recognition and Docking 5 Affects of Water on Ligand Recognition and Docking shift, leading to a 1.1 A r.m.s.d. between docking and crystallographic poses in both cases. Intriguingly, the aldehyde in the docking 15930314 pose of 24 was rotated by,90 degrees, but this variation did not affect the pose prediction. For compound 22, the correspondence between docking and crystallographic pose was slightly worse . In the docking pose, the ligand contacted Gly178 and Met228 via the benzimidazole nitrogens. In the crystal structure, 22 rotated away from Met228 to interact with Leu177 and Gly178 via one benzimidazole nitrogen and the methylamine tail. The second benzimidazole nitrogen hydrogen-bonded with two ordered waters; these water molecules were not modeled in the docking. Finally, the docking prediction for 20 was inconsistent with the crystal structure. In the docking pose, the nitrogen on the imidazo ring interacted with Asp233 while the amine made contact with Gly178. This fragment had two configurations in the electron density, both modeled at 50% occupancy and neither of them resembling the docking pose. Discussion Both bulk and ordered solvent effects play crucial roles in ligand binding, and this has motivated the development of methods to model ordered water molecules in molecular design. Disentangling bulk contributions from those of ordered water molecules, and from the other convoluted terms encountered in biologically relevant target sites, has remained challenging. Because of its simplicity, the ability to determine structures to high resolution, the ability to seek and test new molecules prospectively, and to compare results with an analogous site that is closed to the bulk, the Gateless cavity seems well-suited to testing specific solvent-derived terms in protein ligand binding. Three key observations emerge from this study. First, and contrary to our own expectations, opening the cavity to bulk solvent has no general effect on the relative affinities of cationic and 169939-93-9 biological activity neutral ligands; the former continue to bind much more strongly, with the latter barely measurable. The effects on ligand affinity and binding geometry were context-dependent, and whereas several cationic ligands bound weaker to the opened Gateless cavity than to the analogous closed cavity, one cationic ligand had better affinity, 15340224 as did one neutral ligand. The one common theme only emerges from the structures of the ligand-cavity complexes: those ligands that maintained their interactions with the anchoring aspartate and, at the same time, increased inte