Prediction of high-risk patients by genome-wide copy number alterations from remaining cancer after neoadjuvant chemotherapy and surgery

Abstract

Abstract. In breast cancer, changes of gene copy number were analyzed by cDNA microarray-based comparative genome hybridization using post-treatment archived tissues. Genomic DNA was extracted from 45 surgical specimens after chemotherapy. Informative genes were selected by t-test and were statistically validated by prediction analysis using support vector machine in R package. Fluorescence in situ hybridization and quantitative PCR were performed for validation. We developed three clinical models: comparing good vs. poor prognosis (I), comparing good vs. poor prognosis among poor responders (II) and among good responders (III). 158 gene set (I) differentiated high and low risk of relapse group with 92% accuracy. 51 gene set (II) differentiated good and poor prognosis subgroups among poor responders with 99.9% accuracy. 32 gene set (III) differentiated good and poor prognosis subgroups among good responders with 96% accuracy. This approach has potential applications in the identification of high risk of recurrence after neoadjuvant chemotherapy and surgery.