Development of a robust RNA-based classifier to accurately determine ER, PR, and HER2 status in breast cancer clinical samples

Breast cancers are categorized into three subtypes based on protein expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor-2 (HER2/ERBB2). Patients enroll onto experimental clinical trials based on ER, PR, and HER2 status and, as receptor status is prognostic and defines treatment regimens, central receptor confirmation is critical for interpreting results from these trials. Patients enrolling onto experimental clinical trials in the metastatic setting often have limited available archival tissue that might better be used for comprehensive molecular profiling rather than slide-intensive reconfirmation of receptor status. We developed a Random Forests-based algorithm using a training set of 158 samples with centrally confirmed IHC status, and subsequently validated this algorithm on multiple test sets with known, locally determined IHC status. We observed a strong correlation between target mRNA expression and IHC assays for HER2 and ER, achieving an overall accuracy of 97 and 96 %, respectively. For determining PR status, which had the highest discordance between central and local IHC, incorporation of expression of co-regulated genes in a multivariate approach added predictive value, outperforming the single, target gene approach by a 10 % margin in overall accuracy. Our results suggest that multiplexed qRT-PCR profiling of ESR1, PGR, and ERBB2 mRNA, along with several other subtype associated genes, can effectively confirm breast cancer subtype, thereby conserving tumor sections and enabling additional biomarker data to be obtained from patients enrolled onto experimental clinical trials.