A set of 47 glucose analogue inhibitors of glycogen phosphorylase was investigated using 4D-QSAR analysis. A single significant 4D-QSAR model, having no outliers, was found as a function of alignment and conformational sampling. This 4D-QSAR model consists of six grid cell occupancy descriptors which defines the thermodynamic averaged spatial pharmacophore for this set of analogues. The 4D-QSAR model was validated by aligning it upon the crystal structures of glycogen phosphorylase with some bound inhibitors of the training set. Validation of the 4D-QSAR model was realized by establishing the consistency of the types and locations of the grid cell occupancy descriptors relative to the binding interaction sites of the crystal complex. The loss in binding free energy, due to loss in inhibitor conformational entropy upon binding to the enzyme, was computed and found to be in the 0-2 kcal/mol range for these inhibitors. The "active" conformation of each analogue was also postulated from the 4D-QSAR model.

Glucose analogue inhibitors of glycogen phosphorylase, GP, may be of clinical interest in the regulation of glycogen metabolism in diabetes. The receptor geometry of glycogen phosphorylase b, GPb, is available for structure-based design and also for the evaluation of the thermodynamics of ligand-receptor binding. Free energy force field (FEFF) 3D-QSAR analysis was used to construct ligand-receptor binding models. FEFF terms involved in binding are represented by a modified first-generation AMBER force field combined with a hydration shell solvation model. The FEFF terms are then treated as independent variables in the development of 3D-QSAR models by correlating these energy terms with experimental binding energies for a training set of inhibitors. The genetic function approximation, employing both multiple linear regression and partial least squares regression data fitting, was used to develop the FEFF 3D-QSAR models for the binding process and to scale the free energy force field for this particular ligand-receptor system. The significant FEFF energy terms in the resulting 3D-QSAR models include the intramolecular vacuum energy of the unbound ligand, the intermolecular ligand-receptor van der Waals interaction energy, and the van der Waals energy of the bound ligand. Other terms, such as the change in the stretching energy of the receptor on binding, change in the solvation energy of the system on binding, and the change in the solvation energy of the ligand on binding are also found in the set of significant FEFF 3D-QSAR models. Overall, the binding of this class of ligands to GPb is largely characterized by how well the ligand can sterically fit into the active site of the enzyme. The FEFF 3D-QSAR models can be used to estimate the binding free energy of any new analogue in substituted glucose series prior to synthesis and testing.

The 4D-QSAR model developed for a training set of 47 glucose analogue inhibitors of glycogen phosphorylase, and reported in the previous paper in this issue, was used as a basis for developing virtual high throughput screen, VHTS, models to screen a focused combinatorial virtual library of 225 additional inhibitors. Techniques to develop, evaluate, and apply VHTS models derived from 4D-QSAR models are presented. Application of the VHTS models to screen the virtual library results in the prediction of compounds which bind both more, and less, strongly to the enzyme than the best and worst binders of the training set. Analysis of the binding predictions across the virtual library reveals patterns of structure-activity information that can be useful to design new focused libraries. The possible use of overfit QSAR models, with respect to the training data set, as VHTS models is discussed and explored.

ChemInform is a weekly Abstracting Service, delivering concise information at a glance that was extracted from about 100 leading journals. To access a ChemInform Abstract of an article which was published elsewhere, please select a “Full Text” option. The original article is trackable via the “References” option.

No abstract