Dynamic property is highly correlated with the biological functions of macromolecules, such as the activity and specificity of enzymes and the allosteric regulation in the signal transduction process. Applications of the dynamic property to protein function researches have been discussed and encouraging progresses have been achieved, for example, in enzyme activity and protein–protein docking studies. However, how the global dynamic property contributes to protein–protein interaction was still unclear. We have studied the dynamic property in protein–protein interactions based on Gaussian Network Model and applied it to classify biological and nonbiological protein–protein complexes in crystal structures. The global motion correlation between residues from the two protomers was found to be remarkably different for biological and nonbiological complexes. This correlation has been used to discriminate biological and nonbiological complexes in crystal and gave a classification rate of 86.9% in the cross-validation test. The innovation of this feature is that it is a global dynamic property which does not rely directly on the interfacial properties of the complex. In addition, the correlation of the global motions was found to be weakly correlated with the dissociation rate constant of protein complexes. We suggest that the dynamic property is a key determinant for protein–protein interaction, which can be used to discriminate native and crystal complexes and potentially be applied in protein–protein dynamic rate constants estimations. Proteins 2008. © 2007 Wiley-Liss, Inc.