Breast cancer intrinsic subtypes: a critical conception in bioinformatics

Breast cancers have been uncovered by high-throughput technologies that allow the investigation at the genomic, transcriptomic and proteomic levels. In the early 2000s, the gene expression profiling has led to the classification of five intrinsic subtypes: luminal A, luminal B, HER2-enriched, normal like and basal-like. A decade later, the spectrum of copy number aberrations has further expanded the heterogeneous architecture of this disease with the identification of 10 integrative clusters (IntClusts). The referred classifications aim at explaining the diverse phenotypes and independent outcomes that impact clinical decision-making. However, intrinsic subtypes and IntClusts show limited overlap. In this context, novel methodologies in bioinformatics to analyse large-scale microarray data will contribute to further understanding the molecular subtypes. In this study, we focus on developing new approaches to cover multi-perspective, highly dimensional, and highly complex data analysis in breast cancer. Our goal is to review and reconcile the disease classification, underlying the differences across clinicopathological features and survival outcomes. For this purpose, we have explored the information processed by the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC); one of the largest of its type and depth, with over 2000 samples. A series of distinct approaches combining computer science, statistics, mathematics, and engineering have been applied in order to bring new insights to cancer biology. The translational strategy will facilitate a more efficient and effective incorporation of bioinformatics research into laboratory assays. Further applications of this knowledge are, therefore, critical in order to support novel implementations in the clinical setting; paving the way for future progress in medicine.