Virtual screening, using 3D structural information of a ligand, a protein or a ligand-bound protein structure, has rapidly developed into a useful addition to available 2D ligand-based technologies. Although computationally slower than selections performed on the basis of 2D fingerprints and molecular descriptors, 3D virtual screening utilizes more information regarding the molecular-field requirements for biological activity. In addition, it has the potential to identify entirely new scaffolds and more effectively guide ligand optimization programs. With the pressure from the identification of new targets to screen and the need to reduce time, cost and waste of compounds, combinatorial chemistry and high-throughput screening technologies should effectively integrate 3D virtual screening tools to prioritize reagent or compound selection for synthesis and testing.

In this article, both ligand- and receptor-based VS strategies as implemented in the MIMIC and DOCK programs, respectively, reach nearly 50% consensus on high-ranking compounds from a diverse chemical database. MIMIC may prove more likely to identify compounds that are missed by docking approaches due to the use of a single receptor conformation or imperfect scoring functions. DOCK, on the other hand, is less dependent on the reference structure and enforces complementarity with the receptor in regions not explored by the reference molecule used in similarity approaches. Introducing an additional steric field derived from the protein active site, as in DP-MIMIC, can focus ligand-based 3D screening even further at less computational cost than the use of a full-blown protein-ligand scoring function.

In conclusion, flexible 3D similarity searches provide a generally applicable approach to computational lead discovery, independent of the availability of structural data on the biological target. Its intrinsic dependence on the conformation of the reference molecule suggests, however, that a rigid reference scaffold or its experimentally determined bound conformation is preferably desirable. If structural information on the receptor is available, docking approaches can be applied in lead discovery and especially lead optimization. At the optimization stage, one would prefer to use the maximal amount of information available to focus combinatorial libraries or congeneric series towards optimal fit with the binding site. It is, thus, envisaged that a promising strategy would be to combine the use of ligand- and receptor-based virtual screening techniques, allowing one method to overcome limitations of the other.

0 0

Post a comment