In this paper, we propose a method for the analysis of conservation of domain combinations in proteins. Using the method, we extract conserved domain combinations in Homo sapience proteome and examine their GO term annotations in order to understand the co-evolution of domains in a proteome. Unlike conventional methods, which use co-occurrence frequency for evolutionary analysis of domains, the proposed method measures mutual dependency of domains in proteins as well.??According to the results, domains in Homo sapience proteome turn out to form patterns whose members are highly affiliated to one another. Besides, GO term analysis shows that extracted patterns have a tendency of being associated with molecular functions, and molecular functions are more related with mutual dependency than co-occurrence frequency of domains.??Those results indicate that the proposed method adopting mutual dependency outperforms conventional methods in terms of finding domain combinations conserved through evolution for molecular functional collaboration.