DNA sequencing outperforms RNA sequencing for comprehensive fusion detection in lymphoma diagnostics

Abstract

Abstract Background:： Gene fusions constitute essential biomarkers for diagnosing, prognosticating, and guiding therapeutic decisions in hematological malignancies. While Transcriptomic sequencing (RNA-seq) serves as a well-established modality for comprehensive fusion detection in leukemia, its diagnostic efficacy in lymphoma remains undetermined. Although DNA-based next-generation sequencing (NGS) demonstrates significant potential for identifying lymphoma-associated gene fusions, systematic comparisons between DNA and RNA sequencing methodologies for fusion detection in lymphoma are lacking. This study aims to evaluate the concordance between DNA sequencing(DNA-seq) and RNA-seq in detecting lymphoma-associated gene fusions and to establish an optimized NGS-based diagnostic strategy. Methods: We retrospectively analyzed non-Hodgkin lymphoma (NHL) patients diagnosed at our institution between February 2023 and July 2025 that underwent paired targeted DNA-seq and RNA-seq. RNA-seq data underwent fusion detection using STAR-Fusion and Arriba, while targeted DNA-seq data were analyzed with LUMPY. All candidate gene fusions were manually confirmed by reviewing aligned reads in Integrative Genomics Viewer (IGV). We further assessed the impact of sample type, breakpoint location, partner genes, and sequencing quality on gene fusion detection accuracy. Results: We analyzed paired DNA-seq and RNA-seq data from 108 NHL patients (median age 59 years, range 14 - 93; male:female 63:45) using formalin-fixed paraffin-embedded (FFPE) tissues or bone marrow (BM) aspirates. Among these, 22 NHL cases harbored 25 fusion genes, including recurrent fusions in MYC (n=5), BCL2 (n=7), BCL6 (n=5), CCND1 (n=2), and DUSP22 (n=2), as well as 4 rare fusion genes (MPLKIP::SUGCT, CCND3::IGH, IGL::BCL2L1, TOP2B::ISOC1). DNA-seq detected 23 fusion genes (23/25, 92%), with MPLKIP::SUGCT and TOP2B::ISOC1 undetected due to probe design limitations. In contrast, RNA-seq identified only 11 fusion genes (11/25, 44%) distributed across 8 cases, including MYC (n=3), BCL2 (n=2), BCL6 (n=1), CCND1 (n=1), and all 4 rare fusions. Among these, 10 fusion genes (10/11, 90.9%) harbored intronic breakpoints distributed in 7 cases, with 85.7% (6/7) of these cases being bone marrow (BM)-derived samples. Manual IGV review of RNA-negative cases (n=14) revealed low-level supporting reads missed by bioinformatic pipelines in 6 cases: five with intronic breakpoints (including 3 FFPE samples with poor RNA quality) and one with an intergenic breakpoint suggestive of DNA contamination. The remaining 8 RNA-negative fusion genes, 6 featured intergenic breakpoints with IGH partners, 1 FFPE-derived sample with degraded RNA, and 1 BCL6 intron 1 rearrangement undetected despite adequate RNA-seq quality. Conclusion: Our findings demonstrate that DNA-seq outperforms RNA-seq for detecting gene fusions in lymphoma diagnostics. DNA-based approaches showed particular advantages in identifying intergenic breakpoints, IGH partner fusions, and cases with suboptimal sample quality. These findings support the use of DNA-seq as the primary method for clinical fusion genes screening in lymphoma, while RNA-seq may serve as a complementary tool for cases with predominant intronic breakpoints or when investigating novel/rare fusion events.