How common is the funnel‐like energy landscape in protein‐protein interactions?

The goal of this study is to verify the concept of the funnel‐like intermolecular energy landscape in protein–protein interactions by use of a series of computational experiments. Our preliminary analysis revealed the existence of the funnel in many protein–protein interactions. However, because of the uncertainties in the modeling of these interactions and the ambiguity of the analysis procedures, the detection of the funnels requires detailed quantitative approaches to the energy landscape analysis. A number of such approaches are presented in this study. We show that the funnel detection problem is equivalent to a problem of distinguishing between distributions of low‐energy intermolecular matches in the funnel and in the low‐frequency landscape fluctuations. If the fluctuations are random, the decision about whether the minimum is the funnel is equivalent to determining whether this minimum is significantly different from a would‐be random one. A database of 475 nonredundant cocrystallized protein–protein complexes was used to re‐dock the proteins by use of smoothed potentials. To detect the funnel, we developed a set of sophisticated models of random matches. The funnel was considered detected if the binding area was more populated by the low‐energy docking predictions than by the matches generated in the random models. The number of funnels detected by use of different random models varied significantly. However, the results confirmed that the funnel may be the general feature in protein–protein association.