Robust identification of breast cancer molecular subtypes to refine prognosis

Haibe-Kains B

Since the advent of array-based technology and the sequencing of the human genome, scientists attempted to bring new insights into breast cancer biology and prognosis. Perou et al. highlighted the key molecular differences between breast tumors by identifying sets of co-expressed genes and tumors sharing similar gene expressions (Perou et al, Nature 2000). Several subtypes were identified based mainly on ER and HER2 phenotypes and proliferation. Although these early results were promising, the hierarchical clustering used in the original publications lacked of robustness and was hardly applicable to new data. In order to alleviate these constraints, our group recently introduced robust methods: (i) to identify gene modules, i.e. sets of genes that are specifically co-expressed with genes of interest; (ii) to identify molecular subtypes in breast cancer (Wirapati et al., BCR 2008; Desmedt et al. CCR 2008).

Additionally to these new biological insights, several research groups identified prognostic gene expression signatures. Recent validation studies (Buyse et al. JNCI 2006; Desmedt et al., CCR 2007; Haibe-Kains et al. BMC Genomics 2008) supported the good performance of the early breast cancer prognostic signatures, namely GENE70 (van't Veer et al., Nature 2002), GENE76 (Wang et al., Lancet 2005) and GGI (Sotiriou et al., JNCI 2006).

Combining the identification of molecular subtypes and prognostic gene signatures might improve our understanding of biological phenomena involved in breast cancer prognostication. Interestingly, our group showed that the prognostic value of the early gene signatures, mainly driven by proliferation, is limited to the luminal subtype (ER+/HER2-) (Wirapati et al., BCR 2008; Desmedt et al. CCR 2008). In the other subtypes (i.e. ER-/HER2- and HER2+), immune response and angiogenesis might be highly prognostic.

In this presentation, we will introduce a novel prognostic model which integrates the identification of the breast cancer molecular subtypes and uses a local modeling approach to improve state-of-the-art prognostic gene signatures.