|
|
The main aspect in modeling the physico-chemical properties of the ligands is to use them not only for scoring the solutions, but also for generating them. Another central point is the consideration of the internal degrees of freedom of the ligands.
We use two classes of contributions for scoring: paired intermolecular interactions and overlap volumes. Intermolecular interactions with a potential receptor atom that are plausible for both ligands are paired and contribute a qualitative term to the overall score. Density functions of various kinds, such as electron density, hydrophobicity, hydrogen-bonding donor and acceptor potential and the van-der-Waals volume, are used to characterize the physico-chemical properties of the ligands. We compute overlap volumes of these density functions as further scoring contributions.
Conformational flexibility of the test ligand is modeled by a discrete set of molecular conformations [HCC89]. Therefore we attach to each acyclic single bond a set of energetically favorable torsional angles by a fragment-based assignment process. The QCPE-program SCA is used to calculate a discrete set of low-energy conformations for ring systems. The test ligand is split up into components that contain at most one acyclic single bond and always complete ring systems.
To get a feeling for ligand flexibility, we have prepared pictures of 20 randomly generated conformations of a drug molecule.
Our algorithm is a modification of the flexible docking algorithm FlexX [RKL96]. The method consists of three phases. At first we determine a special fragment, called the base fragment, which serves as an anchor for the subsequent superposition steps. Then we place the base fragment onto the reference ligand. Finally we use an iterative incremental construction procedure to superpose the whole test ligand onto the reference ligand, guided by the scores of the partial solutions. In order to do so, in each iteration we consider a subset of the highest scoring solutions from the previous iteration and attach to them the next component with all its possible conformations. Details of the algorithm are presented in [L97].
Example
Homepage
