Inflammatory breast cancer (IBC) is a rare, aggressive cancer found in all the molecular breast cancer subtypes. Despite extensive previous efforts to screen for transcriptional differences between IBC and non-IBC patients, a robust IBC-specific molecular signature has been elusive. We report a novel IBC-specific gene signature (59 genes; G59) that achieves 100% accuracy in discovery and validation samples (45/45 correct classification) and remarkably only misclassified one sample (60/61 correct classification) in an independent dataset. G59 is independent of ER/HER2 status, molecular subtypes and is specific to untreated IBC samples, with most of the genes being enriched for plasma membrane cellular component proteins, interleukin (IL), and chemokine signaling pathways. Our finding suggests the existence of an IBC-specific molecular signature, paving the way for the identification and validation of targetable genomic drivers of IBC.
Keywords: Breast cancer; IBC; IBC signature; Machine learning; Random forest.
© 2021. The Author(s).