Methodological aspects

All docking calculations were done with the program DOCK 4.0 [21]. Sphere centers, to be used for matching with ligand atoms during docking, were generated in the ligand binding site with SPHGEN [11]. Initial sphere sets were manually reduced using Sybyl 6.5 [84] to yield sets of 20-30 well-dispersed spheres. DOCK 4.0 scoring grids were generated with CHEMGRID [85], using a 0.3 A, grid spacing and a 4r distance-dependent dielectric. The number of accepted bumps between receptor and ligand was set to 3. The minimum anchor size was set to 4 atoms. A maximum of 100 iterations and a convergence threshold of 0.1 were set for chemical score minimization. The scope of the sampling is related to the number of starting orientations that can be tried for each base fragment (anchor) and the number of fragment orientations (seeds) that are kept at each stage of the ligand incremental-construction pro cess. During sampling, the number of starting orientations and the number of seeds were set to 100 and 5, respectively.

All similarity calculations were done with the program MIMIC [76]. Two types of Gaussian-based molecular fields were used to evaluate molecular similarities. An atom-centered steric-volume field (MSV) [86] and an unitedatom point-charge electrostatic potential (MEP) [69] are used to represent the steric and electrostatic features of a molecule, respectively, Although these are rather crude approximations, they are generally adequate to reproduce qualitatively most of the important steric and electrostatic features of a molecule [87-89].

A field-based similarity measure (Zab ) between two molecules is computed as

where FA(r) and FB(r) are the molecular fields associated to the reference molecule, A, and the target molecule, B, respectively. Then, in order to compare similarity values from different pairs of molecules, similarity measures are normalized using a cosine-like similarity index, SAB, defined as

Sab depends on the mutual superposition of molecules, as well as on their conformations. Therefore, assuming a rigid reference molecule, optimization of SAB depends on three translational (t) and three rotational (0) degrees of freedom [76,87,88], as well as on a number of conformational degrees of freedom [20,89] for the target molecule (tb, set of rotatable bonds). Depending on the molecular field used to calculate the similarity, SAB can take values between 0 and 1, for positive definite molecular fields, such as MSV, and between -1 and +1, for non-positive definite fields, such as MEP. Following previous validation studies [87-89], throughout the work a similarity index defined by a weighted combination of MSV and MEP similarity indices in a 2:1 ratio (Sab = 0,67-5 i^1 + 0.33-.Sy was used.

In order to find the optimal superposition between the reference and the target molecules, the orientational space is explored using a spherical systematic search [76]. Basically, the reference molecule is kept fixed and the adapting molecule is systematically placed in a number of orientations around the reference molecule, from which optimization of the molecular similarity, as defined by Equation 2, will proceed following normal gradient-seeking techniques. During sampling, the number of starting orientations and gradient convergence criteria were set to 12 (90-degree search) and 0.01, respectively.

An automatic molecular-size checking with respect to the reference molecule decides whether sampling needs to be further extended for smaller target molecules or limited for much larger target molecules.

Finally, MIMIC has recently also been extended to allow the inclusion of an active-site bump-penalty term to the similarity scoring of the different alignments obtained when sampling. This is simply done by evaluating the asymmetic steric-field similarity between the receptor active site, R, and the target molecule, B, with respect to the target molecule as

The final similarity, , between the reference molecule, A, and a target molecule, B, with the inclusion of a bump-penalty term with the receptor, R, is evaluated as „ = Sab - s^' . At present, the steric penalty is applied only after the MIMIC search has been completed. Therefore, this correction has at most influence on the relative ordering of the alignment solutions if steric clashes are detected in the alignments. Throughout this work, a MIMIC calculation including the active site bump penalty will be referred to as DP-MIMIC (Docking Penalized MIMIC calculation). The DP-MIMIC approach can be considered the complementary implementation of the recently developed SP-DOCK approach (Similarity Penalized DOCK calculation), where a ligand-ligand similarity term is included in the docking energy score [20].

0 0

Post a comment