Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous entity with remarkably variable clinical outcome. Gene expression profiling (GEP) classifies DLBCL into activated B-cell like (ABC), germinal center B-cell like (GCB), and Type-III subtypes, with ABC-DLBCL characterized by a poor prognosis and constitutive NF-kB activation. A major challenge for the application of this cell of origin (COO) classification in routine clinical practice is to establish a robust clinical assay amenable to routine formalin-fixed paraffin-embedded (FFPE) diagnostic biopsies. In this study, we investigated the possibility of COO-classification using FFPE tissue RNA samples by massive parallel quantitative reverse transcription PCR (qRT-PCR). We established a protocol for parallel qRT-PCR using FFPE RNA samples with the Fluidigm BioMark HD system, and quantified the expression of the COO classifier genes and the NF-kB targeted-genes that characterize ABC-DLBCL in 143 cases of DLBCL. We also trained and validated a series of basic machine-learning classifiers and their derived meta classifiers, and identified SimpleLogistic as the top classifier that gave excellent performance across various GEP data sets derived from fresh-frozen or FFPE tissues by different microarray platforms. Finally, we applied SimpleLogistic to our data set generated by qRT-PCR, and the ABC and GCB-DLBCL assigned showed the respective characteristics in their clinical outcome and NF-kB target gene expression. The methodology established in this study provides a robust approach for DLBCL sub-classification using routine FFPE diagnostic biopsies in a routine clinical setting. Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous entity with remarkably variable clinical outcome. Among the many biomarkers investigated so far, only the molecular subtypes by cell of origin (COO) classification, the MYC involved chromosome translocation and TP53 mutation, have been consistently shown to bear prognostic value in the setting of rituximab-containing chemotherapy regimens. The COO-classification by whole-genome expression profiling (GEP) classifies DLBCL into activated B-cell like (ABC), germinal center B-cell like (GCB), and Type-III (unclassified) subtypes, with the ABC-DLBCL characterized by a poor prognosis and constitutive NF-kB activation. 1-5 The original classification was based on similarity of DLBCL gene expression to the activated peripheral blood B cells or normal germinal center B cells by hierarchical clustering analysis. 1 Subsequently, Wright et al 3 identified 27 genes that were most discriminative in their expression between ABC and GCB-DLBCL, and developed a linear predictor score (LPS) algorithm for COO-classification. These seminal works are entirely based on retrospective investigations of fresh-frozen (FF) lymphoma tissues. A major challenge for the application of this COO-classification in clinical practice is to establish a robust clinical assay amenable to routine formalin-fixed paraffin-embedded (FFPE) diagnostic biopsies.Several immunohist...