Meta-Analysis of Diffuse Large B-Cell Lymphoma Gene Expression Identifies Novel and Recurrent Biological Connections,

Abstract 3682

Cell of origin classification of diffuse large B-cell lymphoma (DLBCL) identifies biologically and clinically meaningful subsets. In future these classes may inform distinct treatment. However a meta-analysis to assess consistency of class and gene signature associations across multiple DLBCL gene expression datasets is lacking. Furthermore diagnostic material is routinely available in formalin-fixed paraffin embedded (FFPE) form, and there is limited evidence that for DLBCL FFPE derived RNA can produce gene expression datasets comparable to freshly extracted RNA. Using an FFPE derived data-set we develop a platform robust implementation of the cell of origin classifier. With this classifier we observe comparable separation of survival in distinct data-sets derived from fresh and FFPE material. Classification of six data-sets encompassing a total of >900 DLBCL samples with this implementation establishes very similar patterns of gene expression for each DLBCL class. When tested against 12,000 signatures derived from LLMPP, MSigDB, GeneSigDB and genome wide transcription factor motif analysis of all TRANSFAC and JASPAR matrices, this communality translates into reproducible association of ABC and GCB-DLBCL with distinct molecular pathways, chromosomal regions and transcription factor motifs. This meta-analysis thus confirms the validity of FFPE derived data, identifies the most robust DLBCL class and signature associations and provides evidence for specific transcription factor signatures in both ABC and GCB-DLBCL.