Limitations of Detecting Genetic Variants from the RNA Sequencing Data in Tissue and Fine-Needle Aspiration Samples

Abstract

Background: Genetic profiling of resected tumor or biopsy samples is increasingly used for cancer diagnosis and therapy selection for thyroid and other cancer types. Although mutations occur in cell DNA and are typically detected using DNA sequencing, recent attempts focused on detecting pathogenic variants from RNA. The aim of this study was to determine the completeness of capturing mutations using RNA sequencing (RNA-Seq) in thyroid tissue and fine-needle aspiration (FNA) samples. Methods: To compare the detection rate of mutations between DNA sequencing and RNA-Seq, 35 tissue samples were analyzed in parallel by whole-exome DNA sequencing (WES) and whole-transcriptome RNA-Seq at two study sites. Then, DNA and RNA from 44 thyroid FNA samples and 47 tissue samples were studied using both targeted DNA sequencing and RNA-Seq. Results: Of 162 genetic variants identified by WES of DNA in 35 tissue samples, 77 (48%) were captured by RNA-Seq, with a detection rate of 49% at site 1 and 46% at site 2 and no difference between thyroid and nonthyroid samples. Targeted DNA sequencing of 91 thyroid tissue and FNA samples detected 118 pathogenic variants, of which 57 (48%) were identified by RNA-Seq. For DNA variants present at >10% allelic frequency (AF), the detection rate of RNA-Seq was 62%, and for those at low (5-10%) AF, the detection rate of RNA-Seq was 7% (p < 0.0001). For common oncogenes (BRAF and RAS), 94% of mutations present at >10% AF and 11% of mutations present at 5-10% AF were captured by RNA-Seq. As expected, none of TERT promoter mutations were identified by RNA-Seq. The rate of mutation detection by RNA-Seq was lower in FNA samples than in tissue samples (32% vs. 49%, p = 0.02). Conclusions: In this study, RNA-Seq analysis detected only 46-49% of pathogenic variants identifiable by sequencing of tumor DNA. Detection of mutations by RNA-Seq was more successful for mutations present at a high allelic frequency. Mutations were more often missed by RNA-Seq when present at low frequency or when tested on FNA samples. All TERT mutations were missed by RNA-Seq. These data suggest that RNA-Seq does not detect a significant proportion of clinically relevant mutations and should be used with caution in clinical practice for detecting DNA mutations.

Detection of Multiple Types of Cancer Driver Mutations Using Targeted RNA Sequencing in NSCLC

“…and was consistent with that of a previous study 6 . There were two mutants whose DNA variant frequency was much higher than that observed with the RNA panel.…”

“…Comparing whole-exome sequencing (WES) to whole-transcriptome sequencing in terms of the detection of SNVs in BRAF and RAS genes in tumor tissues, the study showed that 94% of the SNVs (variant allele frequency (VAF) >10%) detected by WES were successfully detected by RNAseq, but only 11% of SNVs (VAF = 5-10%) could be detected by RNA-seq. The results demonstrated that whole-transcriptome analysis with total RNA-seq had a low sensitivity to SNV mutations 6 . In this study, we collected over 1000 clinical NSCLC tissue samples and compared the results from targeted DNA and RNA NGS in parallel.…”

“…Although reports have shown that complex rearrangement with ALK, ROS1 and RET can be detected by DNA-based NGS and most of the detected gene fusions are sensitive to targeted drugs [10][11][12] , the results from DNA-based NGS fail to provide direct evidence to show whether those rare fusions can generate transcribed products with desired functions 13 . Therefore, independent tests, such 6 . In this study, we collected over 1000 clinical NSCLC tissue samples and compared the results from targeted DNA and RNA NGS in parallel.…”

“…For example, RNA testing may 1) have increased alignment errors near RNA splicing sites; 2) have an increased error rate in the process of reversed transcription; 3) fail to detect non-expressed or low-level variations due to different levels of expression; and 4) confuse RNA editing sites with mutations. It has been demonstrated that in tumor tissues RNA sequencing (RNA-seq) of SNVs in BRAF and RAS genes successfully detected 94% of the mutations with a frequency >10% detected by DNA exome sequencing; however, only 11% of the mutations with a variant allele frequency (VAF) of 5-10% were detected by RNA-seq 6 .…”

Identification of 3’-UTR single nucleotide variants and prediction of select protein imbalance in mesial temporal lobe epilepsy patients

“…Detection of variants using RNA-seq data has one more limitation coming from the fact that the RNA-seq data depend on the degree of expression and stability of the mRNA. Further, variant alleles with low frequency values are poorly detected by the RNA-seq method [ 34 ]. Though final confirmation of variant-effect predictions have to come from experiments, computational methods suggest what experiments to do.…”