Protein protein interaction (PPI) is very important information for constructing biological pathways in this systems biology era. Recently many PPI-related databases have been created by high-throughput wet-lab methods. However, in-silico methods developed to predict PPIs are significant techniques for obtaining the whole aspect of PPI networks. Functional regions of a protein defined by specific amino-acid sequences are the key components on determining the role the protein play in a biological process. Association rule mining is a popular data mining skill for finding the association of components in an itemset. Therefore, to mining the associations of functional regions of two interacting proteins will be helpful for PPI prediction. In this study, we collected yeast PPI data from DIP and IntAct, and downloaded the information for functional regions of proteins from Uniprot. A web-based system was designed to integrate, process and mine these data to create some rules based on functional region association. PPIs of other species were used to evaluate these rules. In result, over 80% association rules produced from yeast PPI data in other species. This indicated that the rules learning from known PPI provide good references for PPI prediction.
|Journal||International journal of biomedical soft computing and human sciences|
|Publication status||Published - 2008|