Association of a four-gene decision tree signature with response to platinum-based chemotherapy in patients with triple negative breast cancer.

1006Background: Approaches to capture the molecular complexity of triple negative breast cancers (TNBCs) are lacking. We sought to classify TNBCs into subgroups with common biological features based on transcriptomic and genomic data using a Bayesian algorithm. Methods: Matched gene expression and copy number microarray data was available for the Guy’s (n = 88) and METABRIC (n = 112) TNBC cohorts. CONEXIC was used to derive a decision tree signature for classification. Performance of the signature was tested in 7 TNBC cohorts (total n: 1,368), including 2 clinical trials assessing the efficacy of gemcitabine and carboplatin with and without iniparib. In the early-stage PrECOG 0105 Phase II neoadjuvant trial (n = 43), subtypes were evaluated in relation to response by residual cancer burden (RCB). In the metastatic Sanofi Phase III trial (n = 224), subtypes were assessed by RECIST. Results were compared to the BL1 TNBCtype-4 subtype and assessed using a multivariate analysis. Results: The integrative analy...