Insulin is a crucial protein for biological processes. Insulin protein converts glucose in the blood into energy. If the work of insulin is disrupted it can lead to Diabetes Mellitus [1]. In biological processes, Insulin does not work alone, but is assisted by other proteins that supporting function and activation of insulin. Therefore, to find out what proteins have an influence on insulin, an analysis of Protein-Protein Interactions (PPIs) is needed. Recent developments in high throughput Experimental biology and Computational biology have produced large data protein-protein interactions (PPIs), which are represented as networks, where nodes correspond with proteins and edges correspond to interactions between proteins [2]. Protein interactions have a correlation with protein function, so the equation of protein function forms the interaction between one protein with another protein. It is known that proteins that interact physically tend to be involved in the same cellular processes, and mutations in their genes can cause similar disease phenotypes [3]. The functions of proteins known by analyzing the structure of Gene Ontology (GO)[4]. The same functional between proteins can be measured by semantic similarity, the function that returns numerical values reflects the closeness of meaning between the two ontological terms affixing protein information. GO is a repository of biological ontologies, gene annotations and gene products. Although the annotation data are based on published evidence originating from most unreliable high throughput experiments, they are often used as a benchmark for functional characterization due to their completeness [3][5]. In the research of G. Montanez and Y. Cho assess the reliability of PPI using GO annotation data determined experimentally and concluded computationally. While using the inferred annotation data to trim the inferred protein interactions can be surprising, the resulting bias is in the direction of confirming the validity of PPI, so that true interactions cannot be classified as wrong, at the expense of leaving some fake PPIs undetectable. While acknowledging the disadvantages of such an approach, this allows leveraging freely available GO data to potentially improve the reliability of PPI data sets [6]. GO annotation data is represented as Directed Acyclic Graph (DAG) which only provides information on the relationship of functions to one another. To give a meaningful value to the DAG, the network analysis method can be used for weighting each leaf [7]. The graphical approach has the advantage of determining centralization in the network, and the central node determines an important role in biological processes [8]. Therefore, the centrality method is suitable to be used as a weighting value on DAG to get features on each function of protein molecules. Some researchers use the centrality method to analyze PPI such as Centrality Closeness (CC), Edge Clustering Centrality Coefficient (NC), Intermediate Centralness (BC), Degree Centrality (DC), Eigen Vector Centrality (EC), Information Centrality (IC) and