Fig. 1

Workflow for developing a support vector machine (SVM) BRCA1-like classifier and application to publicly available datasets for biological discovery. In step 1, a new SVM-based BRCA1-like classifier is trained on re-processed and normalized array copy number data. In step 2, the receiver-operation characteristic (ROC) curves were used for evaluating our BRCA1-like classifier in a training and test set (AUC _training = 1.00, AUC _test = 0.75). In step 3, we applied the SVM classifier to tumors in the The Cancer Genome Atlas (TCGA) and Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) data sets. Finally, in step 4, we performed bioinformatics and statistical analyses attempting to understand the biological characteristics of hormone-receptor-positive breast tumors predicted to be BRCA1-like by our SVM classifier