Abstract 328: An integrated analysis of three distinct IBC/non-IBC Affymetrix gene expression data sets to study the transcriptional heterogeneity both between IBC and non-IBC and within IBC

Abstract

Abstract Introduction. Several studies have applied gene expression profiling to inflammatory breast cancer (IBC). Most of these studies were underpowered, mainly because IBC is a rare disease and hence, IBC sample sizes are small. Here, we present an integrated analysis of three distinct gene expression data sets of IBC and non-IBC samples – thus with enhanced power – to further uncover the IBC-specific molecular biology. Materials & Methods. Three Affymetrix gene expression data sets of 137 IBC and 252 non-IBC samples were integrated. Data were normalized, and genes with high signal-to-noise ratios in at least 1% of the arrays were filtered in. The samples were classified according to their molecular subtypes. IBC-specific heterogeneity was investigated using hierarchical clustering, coupled with silhouette score analysis. Supervised analysis, comparing IBC with non-IBC, was performed in non-stage-matched, stage-matched, and molecular subtype-matched approaches with global testing. IBC-specific activated pathways, miRNA-families, and transcription factors were identified with a target gene analysis approach. Results. Four robust IBC sample clusters were identified, clearly associated with the molecular subtypes, with a predominance of the combined Basal-like, ErbB2+, and Luminal B subtypes (∼70% vs. ∼40% in non-IBC; P<0.0001). When we compared IBC to non-IBC, stage-matched and non-stage-matched differences were identified (global test, P<0.0001). After comparing IBC with non-IBC samples within each of the molecular subtypes, differences persisted only within the luminal A and normal-like subtypes (global test, P<0.0001 and P=0.046, respectively). Target gene analysis identified two molecular pathways (TGFβ and INFα), 16 transcription factors (e.g. NKX2.2, FOXM1), and 10 miRNA families that were differentially activated in IBC and non-IBC (FDR<0.01) but not in the four identified IBC sample clusters. Conclusions. This meta-analysis demonstrates that IBC is indeed transcriptionally heterogeneous and that differences between IBC and non-IBC are predominated by the differences in the distribution of the molecular subtypes between IBC and non-IBC. Taking this into account, we were able to more accurately spot IBC-specific changes in pathway, miRNA, and transcription factor activation. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 328. doi:10.1158/1538-7445.AM2011-328