Implementation of a High-Accuracy Targeted Gene Expression Panel for Clinical Care

Abstract

This study describes the validation of a clinical RNA expression panel with evaluation of concordance between gene copy gain by a next-generation sequencing (NGS) assay and high gene expression by an RNA expression panel. The RNA Salah Targeted Expression Panel (RNA STEP) was designed with input from oncologists to include 204 genes with utility for clinical trial prescreening and therapy selection. RNA STEP was validated with the nanoString platform using remnant formalin-fixed, paraffin-embedded-derived RNA from 102 patients previously tested with a validated clinical NGS panel. The repeatability, reproducibility, and concordance of RNA STEP results with NGS results were evaluated. RNA STEP demonstrated high repeatability and reproducibility, with excellent correlation (r > 0.97, P < 0.0001) for all comparisons. Comparison of RNA STEP high gene expression (log2 ratio ≥ 2) versus NGS DNA-based gene copy number gain (copies ≥ 5) for 38 mutually covered genes revealed an accuracy of 93.0% with a positive percentage agreement of 69.4% and negative percentage agreement of 93.8%. Moderate correlation was observed between platforms (r = 0.53, P < 0.0001). Concordance between high gene expression and gene copy number gain varied by specific gene, and some genes had higher accuracy between assays. Clinical implementation of RNA STEP provides gene expression data complementary to NGS and offers a tool for prescreening patients for clinical trials.

Figure 1

Samples tested, assay comparison, and RNA Salah Targeted Expression Panel (STEP) gene list. A: Pie graph of sample cohort 1 tumor types. B: Venn diagram of genes covered in both RNA STEP (204 genes) and STAR next-generation sequencing (NGS; 170 genes). The RNA STEP and STAR NGS panels assess 92 common genes, including 38 (purple) tested for gene copy number variants with STAR NGS. C: List of RNA STEP genes grouped by biological pathways. The 38 (underlined) genes were used to compare the concordance and correlation between assays. EMT, epithelial-mesenchymal transition; GPCR, G protein-coupled receptor; MAPK, mitogen-activated protein kinase; NFKB, nuclear factor kappa-light-chain-enhancer of activated B cells; NK, natural killer; NRF2, nuclear factor erythroid 2-related factor 2; PI3K, phosphatidylinositol 3-kinase; SCLC, small cell lung cancer; VEGF, vascular endothelial growth factor.

Figure 2

Precision of the RNA Salah Targeted Expression Panel assay. Reproducibility of the assay shown as scatterplot of the normalized log2 ratio between technologists (Techs) 1 and 2 (A), nanoString nCounter (NS) instruments 1 and 2 (B), reagent lots 1 and 2 (C), and days 1 and 14 (D).

Figure 3

RNA Salah Targeted Expression Panel (STEP) gene expression results for 38 genes covered for copy number changes by STAR next-generation sequencing (NGS). A: Box-and-whisker plots illustrating gene expression (log2 ratio) results by the RNA STEP assay for 102 samples. Each dot represents the result of one clinical sample for the gene shown in the y axis. The red dots indicate samples with a gene copy number ≥ 5 or MET exon 14 skipping (STAR NGS). The blue dotted lines indicate the log2 ratio cutoffs of 0 and 2. B: Two-by-two table for gene copy number gain by NGS (copy number ≥ 5) versus high gene expression by RNA STEP (log2 ratio ≥ 2). NPA, negative percentage agreement; NPV, negative predictive value; PPA, positive percentage agreement; PPV, positive predictive value.

Figure 4

MET exon 14 skipping concordance and correlation in STAR next-generation sequencing (NGS) and RNA Salah Targeted Expression Panel (STEP). A:MET exon 14 skipping gene expression (log2 ratio) results by the RNA STEP assay for the 102 samples shown in box-and-whisker plots. B: Concordance in both platforms. C: Correlation in both platforms. NPA, negative percentage agreement; NPV, negative predictive value; PPA, positive percentage agreement; PPV, positive predictive value.

Supplemental Figure S1

RNA Salah Targeted Expression Panel (STEP) gene expression results for the 102 clinical samples. The expression (log₂ ratio) of each clinical sample for a particular gene is denoted by a black dot in the box-and-whisker plot. Log₂ ratio 1 and 2 cutoffs are denoted by the blue dotted lines.

Supplemental Figure S2

Comparison of gene copy number by next-generation sequencing (NGS) and gene expression by RNA Salah Targeted Expression Panel (STEP; log₂ ratio). A: Correlation of results from six genes with at least nine paired results and at least five gene copies by STAR NGS; log₂ ratio 1 and 2 cutoffs are denoted by black dotted lines. B: Bar graphs of the Pearson correlation r values from A. Outliers (with low tumor cell percentage (≤30%) or RNA input amount (≤150 ng) were removed in the Pearson correlation graph. N = 7 outliers (B). ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗∗P < 0.0001. ns, not statistically significant.

Supplemental Figure S3

Comparisons of block age, tumor cell percentage, and input amounts. Accuracy, positive percentage agreement (PPA), and negative percentage agreement (NPA) were compared from samples with different specimen block ages (A), different tumor cell percentage (B), and different RNA input amounts (C). N = 63 for >2 years (A); N = 30 for ≤2 years (A); N = 89 for >30% (B); N = 13 for ≤30% (B); N = 97 for ≥150 ng (C); N = 5 for <150 ng (C).

Supplemental Figure S4

Histogram of the Spearman correlation for 191 genes with comparison of RNA Salah Targeted Expression Panel and RNA-sequencing data for 25 squamous cell lung cancer samples. The vertical dashed line represents the mean correlation (0.68).