The diagnostic utility of targeted gene panel sequencing in discriminating etiologies of cytopenia. Issue 10 (7th August 2019)

• Record Type: Journal Article
• Authors: Zheng, Gang; Chen, Ping; Pallavajjalla, Aparna; Haley, Lisa; Gondek, Lukasz; Dezern, Amy; Ling, Hua; De Marchi, Federico; Lin, Ming‐Tseh; Gocke, Christopher
• Is Part Of: American journal of hematology. Volume 94:Issue 10(2019:Oct.)
• Page Start: 1141
• Page End: 1148
• Publication Date: 2019-08-07
• Subjects: Hematology -- Periodicals; 616.15
• DOI: 10.1002/ajh.25592
• Languages: English
• ISSNs: 0361-8609

Abstract

Abstract: The diagnostic utility of somatic mutations in the context of cytopenias is unclear: clonal hematopoiesis can be found in healthy individuals, patients with aplastic anemia (AA), clonal cytopenia of undetermined significance (CCUS) and myelodysplastic syndrome (MDS). We examined a cohort of 207 well‐characterized cytopenic patients with a 640‐gene next generation sequencing (NGS) panel and compared its diagnostic utility with a "virtual" 41 gene panel. The TET2, SF3B1, ASXL1, and TP53 were the most commonly mutated genes (frequency > 10%). Mutations in the 640‐gene panel show high sensitivity (98.3%) but low specificity (47.6%) for diagnosis of MDS. Notably, mutations of splicing factors and genes in the RAS pathway are relatively specific to MDS. Furthermore, high variant allele frequency (VAF) predicts MDS: when the VAF is set at 20%, the positive predictive value (PPV) for MDS is 95.9%, with a specificity of 95.3%. The presence of two or more somatic mutations with ≥10% VAF showed a PPV of 95.2%. While the "virtual" 41‐gene panel showed a mild decrease in sensitivity (95.7% vs 98.3%), 100% specificity was observed when either VAF was set at ≥20% (100% vs 95.3%), or two or more somatic mutations had VAFs ≥ 10%. Our study shows targeted gene panel sequencing improves the diagnostic approach and accuracy for unexplained cytopenia, with its high sensitivity and high PPV for MDS when applying VAF cutoffs. Furthermore, a 41‐gene panel was shown to have at least comparable performance characteristics to the large 640‐gene panel.