Abstract
Computational drug design is highly dependent on the model being used for docking and fitting of structure. Those models are derived mainly from X-ray structures and hopefully represent the physiologically relevant form of the protein in solution. And here is the conundrum: the conditions from which the crystals are derived are usually far from physiological in terms of ionic strength, pH, identity of salts, and presence of other additives, and temperature at which the data were acquired. Even with a molecule bound, the conformations observed may not be the correct ones, partly because of adjustment of the proteins to a molecule not representative of substrate, and partly because there are too many structures of complexes in which the small molecule has been fitted incorrectly. So far, there has been some progress on all of these issues and we can explore ways to address them further, partly at the experimental level and partly at the computational level.