Abstract
Breast cancer is diagnosed in 2.3 million women each year, and kills 685,000 (~30% of patients) worldwide. Breast cancer prognosis for many subtypes has improved due to treatments targeting Estrogen Receptor (ER), Progesterone Receptor (PR), and Human Epidermal growth factor Receptor 2 (HER2). In contrast, patients with triple-negative breast cancer (TNBC) tumors, which lack all three commonly-targeted membrane biomarkers, more frequently relapse and have lower survival due to lack of tumor-selective TNBC treatments. We performed a secondary TNBC analysis of 196 samples across 10 publicly available bulk RNA-sequencing studies to better understand the molecular mechanism(s) of disease and predict robust biomarkers that could be used to improve diagnostic capabilities for TNBC. Our analysis identified ~12,500 significant differentially expressed genes (FDR-adjusted p-value < 0.05) including KIF14 and ELMOD3, and two significantly modulated pathways. Additionally, our novel findings include highly-accurate biomarkers from machine learning methods including CIDEC (97.1% accurate alone), CD300LG, ASPM, and RGS1 (98.9% combined accuracy); as well as TNBC subtype-differentiating biomarkers. We then experimentally and computationally validated a subset of these findings. The results from our analyses can be used to better understand the mechanism(s) of disease and contribute to the development of improved diagnostics and/or treatments for TNBC.