Purpose: Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous disease with distinct molecular subtypes. The most established subtyping approach, the "Cell of Origin" (COO) algorithm, categorizes DLBCL into activated B-cell (ABC) and germinal center B-cell (GCB)-like subgroups through gene expression profiling. Recently developed immunohistochemical (IHC) techniques and other established methodologies can deliver discordant results and have various technical limitations. We evaluated the NanoString nCounter gene expression system to address issues with current platforms.Experimental Design: We devised a scoring system using 145 genes from published datasets to categorize DLBCL samples. After cell line validation, clinical tissue segmentation was tested using commercially available diagnostic DLBCL samples. Finally, we profiled biopsies from patients with relapsed/refractory DLBCL enrolled in the fostamatinib phase IIb clinical trial using three independent RNA expression platforms: NanoString, Affymetrix, and qNPA.Results: Diagnostic samples showed a typical spread of subtypes with consistent gene expression profiles across matched fresh, frozen, and formalin-fixed paraffin-embedded tissues. Results from biopsy samples across platforms were remarkably consistent, in contrast to published IHC data. Interestingly, COO segmentation of longitudinal fostamatinib biopsies taken at initial diagnosis and then again at primary relapse showed 88% concordance (15/17), suggesting that COO designation remains stable over the course of disease progression.Conclusions: DLBCL segmentation of patient tumor samples is possible using a number of expression platforms. However, we found that NanoString offers the most flexibility and fewest limitations in regards to robust clinical tissue subtype characterization. These subtype distinctions should help guide disease prognosis and treatment options within DLBCL clinical practice.