Chemistry Reference
In-Depth Information
programs being developed there has been a growing need for appropriate evaluations of these programs focused on
virtual screening, pose prediction or various combinations. Metrics, standardized data sets and statistical analysis
can be useful for evaluating real performance differences between docking programs.
In recent work, PhDOCK, ICM, GLIDE, FlexX, DOCK and SURFLEX were evaluated by comparing their utility in
selecting active compounds from a database of decoys as well as generating and identifying docking poses which are
close to the X-ray conformation [77]. Despite its limitations the Directory of Useful Decoys (DUD) can be used to
evaluate the performance of docking programs in virtual screening and remains a valuable virtual screening
benchmark available to the entire chemistry community [263].
Among the docking methods used by the workers, the flexible anchor-and-grow algorithm was used in DOCK 6.1,
base fragment placement and incremental construction was used in FlexX V2.03, an algorithm with precomputed
grids occurring in a hierarchical fashion was used in GLIDE v4.5, optimization of flexible ligands by internal
coordinates in a grid-based receptor field was used by ICM v3.5-1, 3D pharmacophores simultaneously docked into
the target binding site and then scored using the contact scoring function was used in PhDOCK, and empirical
Hammerhead scoring function using an idealized active site ligand to generate ligand poses by incremental
construction and a crossover procedure that combines pieces from distinct poses was used in the Surflex algorithm.
The ability of a docking program to reproduce a ligand pose close to that found in an X-ray complex is often a
critical determinant of the program's effectiveness for structure-guided design. Analysis of the softwares indicated
general trends in accuracy that are specific for particular protein families. Modifying basic parameters in the
software was shown to have a significant effect on docking and virtual screening results, suggesting that expert
knowledge is critical for optimizing the accuracy of these methods [77].
There was recently [404] reported an assessment of some programs for ligand binding affinity prediction. FLEXX,
X-Score, AutoDock and BLEEP were examined for their performance in binding free energy prediction in various
situations including co-crystallized complex structures, cross docking of ligands to their non-cocrysallized receptors,
docking of thermally unfolded receptor decoys to their ligands, and complex structures with 'randomized' ligand
decoys. There was not found a satisfactory correlation between the experimental estimated binding free energies
over all the datasets tested. A strong correlation between ligand molecular weight-binding affinity correlation and
experimental predicted binding affinity correlation was found.
In order to identify receptor-ligand interactions through an
ab inito
approach workers have recently demonstrated a
qualitative relation between the electric characteristics and binding affinity of a complex-receptor-ligand; a large
binding affinity correlates with a large charge transfer which allows analysis of binding interactions of complexes
using small computational resources with acceptable reliability of the results [402].
Largescale applications of high-throughput molecular mechanics with Poisson-Boltzmann surface area (PBSA) for
routine physics-based scoring of protein-ligand complexes were made [403]. Statistically significant correlation was
observed with experimentally measured potencies. The calculations illustrate the feasibility of procedural
automation of physics-based scoring calculations to produce ordered binding-potency estimates for protein-ligand
complexes with sufficient throughput for realization of practical implementation into scientist workflows in an
industrial drug discovery research setting.
A new alternative method for the evaluation of docking performance was recently reported: RSR vs RMSD, i.e an
assessment criterion for docking poses in which experimental electron density is taken into account when evaluating
the ability of docking programs to reproduce experimentally observed binding modes. Three docking programs
(Gold, Glide, and Fred) were used to generate poses for a set of protein-ligand complexes for which the crystal
structure is known. The new criterion is based on the real space R-factor (RSR) which measures how well the ligand
fits the experimental electron density by comparing that density to the expected density, calculated from the
predicted ligand pose. The RSR-based measure is compared to the traditional criterion, the root-mean-square
distance (RMSD) between the docking pose and the binding configuration in the crystallographic model. The results
highlight several shortcomings of the RMSD criterion that do not affect the RSR-based measure. The RSR-derived
approach allows a more meaningful
a posteriori
assessment of docking methods and results [405-410].
