The protein-protein interaction network (PPIN) is essential for functional processing and mechanism of multiple complex diseases. Recently, control theory has applied to protein interaction with the aims of identify the minimum set of nodes that can drive the whole network to the desired state. Here, we use different statistic network inference methods to generate the highest-scored re-ranking gene list as the source for constructing protein-protein interaction network. Then we characterize structural controllability of directed and weighted PPINs for breast cancer stages. The maximum matching approach for controllability analysis allows classifying nodes into three categories: critical, intermittent and redundant. This leads to identifying the most important proteins as critical nodes for each stage of breast cancer. In total, 70 critical nodes as drug targets have been revealed across stages in this study.