Studies of protein movements and their accounting in docking
To perform receptor-based virtual screening, one usualy use single receptor structure model (in most cases obtained from crystalyzed complex with native ligand) to dock ligand database. But in such model the protein active site is structuraly adopted only for native ligand or its analogues. As a consequence, the uncorrect estimation of binding energy can take place in case we dock ligands that significantly differs from native. Therefore, it negatively affects docking scores and further selection of potent ligands.
To improve docking results, it is rationaly to use multiple "rigid" receptor conformations derived from molecular dynamics simulation or X-ray (or NMR) data. Here we show, that using only two different receptor conformations has significant advantages over traditional “single-conformation” docking.
DOCK 4.0 and TOPBUILDER packages has been used to dock 2000 4-aminoquinoline and 4-aminoquinazoline derivatives. Docking procedure was performed in two ways. In one case we used spatial model of CK2 crystal structure obtained from Brookhaven Protein Data Bank (PDB, complex with ATP) to dock ligand database. In second case we used two CK2 conformations derived from MD calculations in GROMACS.
Docking in crystal structure allowed to find 3 active ligands (IC50 10-6). On the contrary, docking in MD-derived receptor structures results in 6 active ligands (IC50 10-6)
Correlation of the obtained results are represented on Fig. 3.
Fig.3 Comparison of the results obtained by MD-docking approach and docking into CK2 crystal structure. The used CK2 conformations are schematically represented as rectangles.
This study was indicate, that some receptor conformations more favorable to dock certain ligands. In particular, the active ligands identified with “multireceptor” docking has not been identified with docking in single crystal structure.
Nevertheless, using such approach we faced with some important problems:
• Generating of multitude of receptor conformations that most fully describe receptor movements
• Selecting the only one most optimal receptor conformation for docking of each ligand in order to increase docking/scoring accuracy and decrease compupational expences .
To carry out the task of selection of most optimal receptor structures from MD trajectory or PDB, we develop SELECTOR software. The selection realized accounting structural features of each tried ligand. At the time we focused on parametrization and testing of the SELECTOR.
© YARMOLUK RESEARCH LAB
