P004 | COMPARATIVE ASSESSMENT OF TARGETED NGS PANELS AND ALGORITHMS FOR OPTIMAL MOLECULAR SUBTYPING OF DLBCL IN CLINICAL PRACTICE

Molecular classification of Diffuse large B cell lymphoma (DLBCL) is recognized as a valuable tool for prognostication and therapeutic decision-making, enabling the identification of patients at higher risk and informing the selection of targeted drugs. While probabilistic approaches based on whole exome sequencing (WES) or targeted next generation sequencing (NGS) have been proposed for molecularly classify DLBCL in the subclusters MCD, EZB, N1, BN2, ST2, and A53, the clinical utility and performance of commercially-available NGS panels remain to be fully established. Here, we assessed two targeted NGS panels commonly used in hematopathology laboratories in Italy, respectively comprising 54 and 73 genes that encompass frequent mutations in B and T lymphomas. The performance of these panels was evaluated against established classifiers as LymphPlex (targeted NGS used in the ongoing prospective, phase 3, randomized GUIDANCE-2 trial NCT0535134) and LymphGen (WES-based). Analyses were restricted to genes shared between the panels and the classifiers and concordance was measured using Cohen’s Kappa and confusion matrices in two large, clinically annotated DLBCL cohorts (n=574 and n=664). The 73-gene panel demonstrated substantial agreement with LymphPlex showed (Kappa 0.786), showing high sensitivity and specificity for the EZB (97.7% and 98.5%) and MCD (86.7% and 98.4%) subtypes, but lower sensitivity for BN2 (31.9%) and ST2 (25.0%) despite high specificity (100% and 99.6%, respectively). The 54-gene panel yielded similar results (Kappa 0.731), with slightly reduced sensitivity for EZB (95.6%) and MCD (73.5%), and specificity consistently above 98%. In contrast, agreement between the 73-gene panel and LymphGen, the agreement was only fair (Kappa 0.245), with marked reduction in sensitivity for most subtypes (e.g., 23.1% for EZB and 38.1% for MCD), although specificity remains high. The 54-gene panel showed even lower concordance with LympGen (Kappa -0.0303), with sensitivities of 21.1% for EZB and 20.3% for MCD, but specificity above 90% for most classes. These findings provide clinically valuable information for integrating molecular profiling in the DLBCL diagnostics. They also highlight the potential for targeted NGS panels to identify strongest molecular predictors, supporting the future development of optimized clinical-scale panels for risk stratification and personalized therapy in DLBCL.