Protein exposes its specific function by interaction with other proteins. Thus the function of the protein whose function is not known yet can be inferred from interacting protein because the interacting proteins may express similar functions. Protein-protein interaction occurs by physical interactions of protein domains. Therefore, the domain-domain interaction data can be used for predicting protein-protein interaction. In this paper, we select the weighted frequency scoring method for predicting domain-domain interaction and evaluate it. The domain-domain interaction is derived from established protein-protein interaction and protein-protein data. The former is taken from existing protein-protein interaction databases and the latter is obtained by applying protein sequences to a protein-domain database, Interpro. Based on the domain-domain interaction data, we develop an interaction prediction system accessible through the web. We design service to deal with bulk data at once and find out the domain interaction information of proteins, domains and sequences automatically. To unify various accession numbers in the database, a table to contain them is built for the protein and the domain respectively. The database is composed of several related tables: protein information, protein-protein interaction, protein-domain, domain-domain interaction and domain information. The services provided by the system will be useful for obtaining the outline of interactions before searching details about each item in prescreening step.