Conclusions

A procedure to determine hydrophobic regions on the surface of a protein and a continuum electrostatic approach for the accurate and efficient docking of a fragment library have been presented. The hydrophobicity maps allow easily to discriminate between hydrophobic and hydrophilic surface regions that are close in space. This was illustrated for the thrombin-NAPAP and p38 MAP kinase-triarylimidazole inhibitor complexes, whose hydrophobi-city maps can be used for de novo design and lead modification. Furthermore, for the thrombin-NAPAP and farnesyltransferase-farnesylpyrophosphate complexes it was shown previously that existing approaches based on the analysis of the surface curvature and/or the electrostatic potential are not as valuable in distinguishing regions where nonpolar and polar groups can bind [ 1]. The fraction of the most hydrophobic receptor regions that are buried at the binding interface is in general particularly high, suggesting that hydro-phobic association is determinant in protein-ligand binding. This confirms previous findings [6,7].

A number of very efficient docking programs have been published recently. Most of them use either a scoring function with a crude approximation of solvation [64-66] or a vacuum energy derived from a molecular mechanics force field [67,68]. The program DOCK, which pioneered the use of geometric criteria to select ligands which best complement the shape of the receptor site [69,70], has been supplemented by the evaluation of lig-and desolvation [71,721. To efficiently screen large databases ofcompounds, DOCK assumes that every ligand desolvates the receptor equally and that the ligand is completely desolvated upon binding. The continuum electrostatic approach implemented in SEED does not make these assumptions.

There are two main advantages in SEED with respect to the multiple copy simultaneous search method (MCSS) which is a force field-based approach for determining optimal positions and orientations of functional groups in a protein binding site [73-78]. These are the inclusion of electrostatic solvation and the determination of all favorable binding modes. The effects of the solvent are neglected in MCSS which calculates the protein-fragment interactions with a vacuum potential [21]. This choice in MCSS was based on the principle that fast methods are necessary to perform effective searches of the binding site and that good candidate ligands subsequently can be ranked in terms oftheir binding energy. A possible difficulty is that minimized positions may be missed or misplaced due to the lack of a solvation correction during the MCSS minimization. The best energy minima without solvation do not always turn out to be those of most interest [76]. Particularly problematic is the docking of apolar fragments which, without inclusion of solvation, are positioned in both hydrophilic and hydrophobic regions of the binding site. This problem is solved in SEED by the prioritization of apolar regions on the protein surface according to low electrostatic desolvation and favorable van der Waals interactions, as well as the efficient use during docking of a protein desolvation look-up table.

SEED samples optimal binding modes and can also find positions which do not necessarily correspond to local minima of the energy function (e.g., a favorable region with relatively flat potential energy in between two well pronounced minima). This is an advantage with respect to MCSS because not all of the molecular fragments, even in potent inhibitors, have optimal interactions with the protein [76].

Fragments are docked as rigid bodies by SEED. For larger ligands with ro-tatable bonds, conformational flexibility can be taken into account by docking different conformations. Programs are available for the automatic generation of diverse low-energy conformations of small molecules [79,80]. In the case of large ligands with many rotatable bonds one could use SEED to find optimal positions for the rigid moieties and then use other techniques which allow for full ligand and eventually also protein flexibility [81].

0 0

Post a comment