Comparing NGS and NanoString platforms in peripheral blood mononuclear cell transcriptome profiling for advanced heart failure biomarker development

• Published in: Journal of biological methods Vol. 7; no. 1; p. e123
• Main Authors: Bondar, Galyna, Xu, Wenjie, Elashoff, David, Li, Xinmin, Faure-Kumar, Emmanuelle, Bao, Tra-Mi, Grogan, Tristan, Moose, Jim, Deng, Mario C
• Format: Journal Article
• Language: English
• Published: United States Journal of Biological Methods 03.01.2020
• Subjects: Benchmark; NanoString; biomarker; advanced heart failure; next generation sequencing; prediction test
• ISSN: 2326-9901; 2326-9901
• DOI: 10.14440/jbm.2020.300

In preparation to create a clinical assay that predicts 1-year survival status of advanced heart failure (AdHF) patients before surgical/interventional therapies and to select the appropriate clinical assay platform for the future assay, we compared the properties of next generation sequencing (NGS) used in the gene discovery phase to the NanoString platform used in the clinical assay development phase. In 25 AdHF patients in a tertiary academic medical center from 2015 to 2016, PBMC samples were collected and aliquoted for NGS RNA whole transcriptome sequencing and compared to 770 genes represented on NanoString's PanCancer IO 360 Gene Expression research panel. Prior to statistical analysis, NanoString and NGS expression values were log transformed. We computed Pearson correlation coefficients for each sample, comparing gene expression values between NanoString and NGS across the set of matched genes and for each of the matched genes across the set of samples. Genes were grouped by average NGS expression, and the NanoString-NGS correlation for each group was computed. Out of 770 genes from the NanoString panel, 734 overlapped between both platforms and showed high intrasample correlation. Within an individual sample, there was an expression-level dependent correlation between both platforms. The low- . intermediate/high-expression groups showed NGS average correlation 0.21 . 0.58-0.68, respectively, and NanoString average correlation 0.07-0.34 . 0.59-0.70, respectively. NanoString demonstrated high reproducibility ( > 0.99 for 100 ng input), sensitivity (probe counts between 100 and 500 detected and quantified), and robustness (similar gene signature scores across different RNA input concentrations, cartridges, and outcomes). Data from NGS and NanoString were highly correlated. These platforms play a meaningful, complementary role in the biomarker development process.