The development of bioinformatics has depended on the contributions of many experts in the various disciplines, such as biologists, chemist, computer scientists, and mathematicians. One of the most widely discussed cases in bioinformatics is the protein grouping. Proteins work together each other to regulate a biological process. From computer science point of view, the interactions that occur between proteins will form a graph, and a mechanism of grouping is done by a unique process, namely clustering. Clustering will be done based on graph of protein interactions. Therefore, this study discusses about a new method that includes in graph clustering. The mechanism is made based on the similarity of protein pairs. If a pair of proteins has high similarity then they will be in the same cluster, and vice versa. As evaluation, this method is implemented on a network of protein domain and compared with grPartition, a well known method for graph clustering. The results show that gCLUPS has a better performance than grPartition in connectivity and separation but not homogeneity.