Individualized survival and treatment response predictions for breast cancers using phospho-EGFR, phospho-ER,...

The development and progression of breast cancer involves the activation of numerous protein kinases, and the change in phosphorylation is a hallmark of protein kinase activation. In this study, we identified a comprehensive profile to predict individual breast cancer patients' survival and treatment responses using the Random Committee algorithm. The profile incorporated a subset of phosphorylated signal protein expressions and several selected clinical factors of breast cancer. The parameters of our profile were identified by supervised feature selection algorithms, Gain Ratio Attribute Evaluation and Relief. The results showed that the overall accuracy of survival prediction reached 92.3% for individual breast cancer patients with the use of the expression profiles of phospho-EGFR, phospho-ER, phospho-HER2/neu, phospho-IGFIR/In, phospho-MAPK, and phospho-p70S6K plus the selected clinical factors. The results also indicated that the overall accuracy of treatment response prediction was 92.6% with the use of the level of phospho-EGFR, phospho-ER, phospho-HER2/neu, phospho-MAPK, and phospho-p70S6K plus the selected clinical information. The prediction system combines multiple signal protein activation profiles and relevant clinical information, and provides a unique guideline to aid individualized decision-making in the clinical management of breast cancer.