RNA Expression Classifiers from a Model of Breast Epithelial Cell Organization to Predict Pathological Complete Response in Triple Negative Breast Cancer

Abstract

Pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) is correlated with better outcomes for breast cancer, especially for triple negative breast cancer (TNBC). We developed RNA expression classifiers from a model of breast epithelial cell organization to predict which patients will achieve pCR to NAC, and which will have residual disease (RD). An exclusive collection of retrospective formalin-fixed, paraffin-embedded (FFPE) pretreatment biopsies from 222 multi-institutional breast cancer patients treated with NAC, including 90 TNBC patients, were processed using standard procedures. A novel strategy using machine learning algorithms and statistical cross-validation were used to develop predictive classifiers based on AmpliSeq differential gene expression analysis of patient samples. Two RNA expression classifiers of 18 genes and 15 genes applied sequentially to the total cohort, classified patients into three distinct classes which accurately identified 83.75% of pCR and 86.62% of RD patients in the total population, and 92.10% of pCR and 80.77% of RD patients in the TNBC subset. This new approach identified a subset of TNBC patients predicted to have RD showing significantly higher levels of Ki-67 expression and having significantly poorer survival rates than the other TNBC patients. Stratification of patients may allow identification of TNBC patients with the worst prognosis prior to NAC, allowing for personalized treatments with the potential to improve patient outcomes.