We report results for the in-silico screening of a database of 10000 flexible compounds against various crystal structures of the thymidine kinase receptor complexed with 10 known inhibitors. The ligands were docked with the stochastic tunneling method using a first-principle based scoring function. Using a rigid receptor we find large deviations in the rank of the known inhibitors depending on the choice of receptor conformation. We then performed a screen in which critical receptor sidechains were permitted to change conformation and found improved scores for those inhibitors that did not dock well in any of the previous screens. These data demonstrate that the failure to dock did not originate from deficiencies in the scoring function or the docking algorithm, but from the neglect of receptor degrees of freedom.