Abstract
Statistical analysis of protein-protein interfaces in a database of pure peptide crystals shows that the distribution of the contact area contains two components: a major exponential distribution and a minor flat distribution. Analysis of two sub-databases provides evidence that the two components represent specific and non-specific contacts, respectively. The probability of an interface with a given area being specific can be estimated. A scaled quantity (contact ratio) is introduced that is more useful than contact area for discriminating specific and non-specific contacts in protein crystals.
Keywords: crystal structure, interface, statistical analysis, specificity of interface