Quantitative prediction of tumor response to neoadjuvant chemotherapy in breast cancer: novel marker genes and prediction model using the expression levels.

BACKGROUND: In breast cancer, the identification of accurate predictors of tumor response to neoadjuvant chemotherapy is of key importance, but none of the critical markers have been validated to date. We attempted to identify potent marker genes genome-wide, and we developed a prediction model for individual response to epirubicin (EPI)/cyclophosphamide (CPM) combination chemotherapy (EC).
METHODS: From 10 human breast cancer cell lines, genes whose expression levels correlated with cytotoxicities of EPI and CPM were chosen through comprehensive gene expression analysis followed by correlation-confirmation study of the quantified expression levels analyzed by real-time reverse transcription polymerase chain reaction (RT-PCR).
RESULTS: We finally selected a total of 4 genes (ANXA1 and PRKCA for EPI; DUSP2 and SERPINA3 for CPM) as reliable prediction markers. Using quantified expression data of genes in 18 tumor samples, we performed multiple linear regression analysis to establish the best linear model that could convert the quantified expression data to show tumor response to the EC therapy (the ratio of tumor size to the baseline, %). Outliers were identified by referring to the value of AIC (Akaike's information criterion) for each sample (AIC/sample) or checking residuals graphically. The multiple linear regression analysis of the selected genes yielded 2 highly predictive formulae for the tumor response: one used all of the genes except SERPINA3 (R = 0.8348, AIC/sample = 4.9182) and the other used all of the 4 genes (R = 0.8224, AIC/sample = 5.0730).
CONCLUSIONS: A study to validate the predictive values of the selected 4 genes is now planned, along with research to determine their functional roles.