Using Outliers to Find Genes Relevant for Basal Breast Cancer

The medical community commonly accepts four subtypes of breast cancer: basal, her2, luminal A, and luminal B. Among them, the basal subtype is the one with the worst prognosis. Quick diagnosing basal subtype and finding important genes related to it is crucial to properly and effectively treat patients. We propose a strategy to identify relevant genes for the basal subtype by finding consistent outliers in the RNA gene expression of basal and non-basal samples. We understand that a single set of genes (e.g., PAM50) to study all the subtypes may not suffice to understand each subtype’s particularities. We compare the relevance of the subset of genes we find to describe the basal subtype against PAM50. In our first experiments, we found 30 genes that correctly clustered 100% the basal samples alone in a single cluster, compared to the PAM50, which grouped several non-basal samples with the basal ones.