Knowing the functional or energetical coupling between two residues in a protein is important for extending our understanding of the protein regulation and functional mechanism. A new method of predicting such couplings between residues from multiple sequence alignment, named Evolutionary Conserved Preference Analysis (ECPA), is proposed in this thesis. The proposed method utilizes the fact that if a certain physical interaction network is functionally or structurally important to a protein, the interaction intensity in that network must be conserved in the evolutionary process. Performance of the method was tested and compared with a previous method, Statistical Coupling Analysis (SCA), by applying it to four known protein residues coupling data sets. The result of the test revealed that; i) in PDZ data set and C2H2 zinc finger data set, our ECPA and SCA method detected the highest coupled interaction relations; ii) in SNase data set and KSI data set, ECPA could detect energetically coupled interaction missed by SCA method. These results indicate that the evolutionary conserved preference is a reliable measure to detect energetically and functionally coupled interactions, and ECPA method and SCA can be used complementarily.