There is a growing interest in using efficient shape complementarity docking algorithms to analyze protein-protein interactions at a large scale. We have realized complete cross-docking of several tens of enzyme/inhibitors proteins. On the one hand, we demonstrate that docking score distributions for the known complexes are not distinguishable from those for the non-interacting pairs. On the other hand, we show that the knowledge of the experimental interfaces applied to the docking conformations permits to retrieve true interaction partners with high accuracy. We further identify the determinants of the molecular recognition between true interactors compared to non-interacting proteins.
展开▼
