Purpose: We sought to systematically define determinants of the response to neoadjuvant chemotherapy to elucidate predictive biomarkers for breast cancer.
Experimental Design: An unbiased systematic analysis was performed in multiple independent datasets to define genes predictive of complete pathologic response (pCR) following treatment with neoadjuvant chemotherapy. These genes were interrogated across estrogen receptor (ER)-positive and ER-negative breast cancer and those in common across three different treatment regimens were analyzed for optimal predictive power. Subsequent validation was performed on independent cohorts by gene expression and IHC analyses.
Results: Genes that were highly associated with the response to neoadjuvant chemotherapy in breast cancer were readily defined using a computational method ranking individual genes by their respective ROC. Such predictive genes of the response to taxane-associated therapies were strongly enriched for cellcycle control processes in both ER-positive and ER-negative breast cancer and correlated with pCR. However, other genes that were specifically associated with residual disease were also identified under other treatment conditions. Using the intersection between treatment groups, nine genes were identified that harbored strong predictive power in multiple contexts and validation cohort. In particular, the nuclear oncogene DEK was strongly associated with pCR, whereas the cell surface protein BCAM was strongly associated with residual disease. By IHC staining, these markers exhibited potent predictive power that remained significant in multivariate analysis.
Conclusion: Systematic computational approaches can define key genes that will be able to predict the response to chemotherapy across multiple treatment modalities yielding a small collection of biomarkers that can be readily deployed by IHC analyses.
ASJC Scopus subject areas
- Cancer Research