Exploiting expression patterns across multiple gene isoforms to identify radiation response biomarkers in early-stage breast cancer patients.

Abstract

In an effort to understand the underlying biology of radiation response along with whole transcriptome effects of preoperative radiotherapy in early-stage breast tumors, we propose two efficient score-based statistical methods that exploit gene expression patterns across all available gene transcript isoforms and identify potential biomarkers in the form of differentially expressed genes and differentially enriched gene-sets. We demonstrate the effectiveness of these two methods using extensive simulation studies that show that both of our methods give improved performance, in terms of statistical power, over the most commonly used methods. By exploiting radiation-induced changes in all available gene transcript isoforms, we identified several statistically significant differentially expressed genes related to PI3K-AKT and JAK-STAT signaling pathways along with radiation-induced oncogenic signaling pathways and tumor microenvironment gene signatures that could be potential targets to improve response to radiotherapy in breast tumors.