Breast cancer is one of the most common malignant tumours. However, the existing functional knowledge about the proteins related to the breast cancer is too general to promote the following study. To find their finer functions, we suggest using the function-specific interaction sub-networks by integrating the functional knowledge of Gene Ontology and the protein-protein interaction network. The results show that the proposed approach is able to reliably find finer functions for these proteins with high precision. The predicted finer functions are highly valuable for guiding the follow-up wet-lab re-search to study the molecular mechanism of breast cancer.