ABSTRACT
Emerging therapies targeting the molecularly distinct GCB and non-GCB/ABC subtypes of diffuse large B-cell lymphoma (DLBCL) have created the need to develop an accurate subtyping assay for routine use. We investigated the potential of QuantiGene Plex (QGP)-branched DNA signal amplification assay-for DLBCL subtyping. We performed in silico analysis of public DLBCL datasets to develop and validate a naïve Bayes classifier, and migrated the resulting 21-gene classifier to QGP and real-time quantitative PCR (qPCR) assays. Forty DLBCL formalin-fixed, paraffin-embedded tumors of known subtype (20 per subtype by gene expression profiling of paired fresh-frozen tissues) were reclassified, and results for QGP (on 38/40 for 21/21 targets) and qPCR (on 40/40 samples for 19/21 targets) compared for recapitulation of microarray data and classification accuracy. The 21-gene bayesian classifier achieved mean area under the curve values >0.9 on independent validation. QGP showed a higher correlation with microarray data (mean R(2) = 0.66 ± 0.05 versus 0.34 ± 0.07; P < 0.0001) and classification accuracy (92.1% versus 78.9%). The proportion of validated targets was also higher for QGP (85.7% versus 47.4%). The QGP protocol was rapid and simple to perform, at a cost similar to qPCR. These promising preliminary results strongly support ongoing work to develop a QGP companion diagnostic assay for DLBCL subtyping.