Exome sequencing has higher diagnostic yield compared to simulated disease-specific panels in children with suspected monogenic disorders

Abstract

As test costs decline, whole-exome sequencing (WES) has become increasingly used for clinical diagnosis, and now represents the primary alternative to gene panel testing for patients with a suspected genetic disorder. We sought to compare the diagnostic yield of singleton-WES with simulated application of commercial gene panels in children suspected of having a genetically heterogeneous condition. Recruitment, singleton-WES and phenotype-driven variant analysis was completed for 145 paediatric patients. At recruitment, clinicians were required to propose commercial gene panel tests as an alternative to WES and nominate a phenotype-driven candidate gene list. In WES-diagnosed children, three commercial options for each proposed panel were identified and evaluated for hypothetical diagnostic yield assuming 100% analytical sensitivity and specificity. We compared the price of WES with the least costly panel in WES-diagnosed children. In WES-undiagnosed children, we evaluated the exonic coverage of their phenotype-driven gene list using aggregate data. WES diagnoses were made in genes not included in at least one-of-three commercial panels in 42% of cases. Had a panel been selected instead, 23% of WES-diagnosed children would not have been diagnosed. In 26% of cases, the least costly panel option would have been more expensive than WES. Evaluation of WES coverage found that at the most stringent level of 20× coverage, the likelihood of missing a clinically relevant variant in a candidate gene list was maximally 8%. The broader coverage of WES makes it a superior alternative to gene panel testing at similar financial cost for children with suspected complex monogenic phenotypes.

Fig. 1

Overview of study methodology and results of simulated panel analysis in WES-diagnosed children who had a viable alternative panel test option proposed. The proposed panels from three commercial providers were assessed in a simulated analysis of those who received a diagnosis by WES (n = 78). In the 57/78 (73%) WES-diagnosed children who had a panel proposed, 13/57 (23%) would have gone undiagnosed by any of the three panels proposed; 11/57 (19.3%) would have gone undiagnosed by at least one of the proposed commercial panels; and 33/57 (57.9%) would have been diagnosed by any of the three proposed panels. A panel was deemed to be diagnostic if it covered the WES-identified gene. The diagnostic yield of panel testing in the WES-undiagnosed group (n = 67) is unknown, but we sought to determine the analytical validity of WES by in silico assessment of coverage of individual phenotype-driven candidate gene lists (Fig. 2)

Fig. 2

Fraction of clinically relevant loci in WES-undiagnosed children’s candidate gene lists falling below three thresholds of coverage quality. At least 92% of potentially pathogenic (clinically relevant) loci in all undiagnosed children’s gene lists were covered by 20 reads or more. At least 96% of potentially pathogenic loci were in regions with >10× coverage. Each bar represents an individual child’s gene list. Note that in the first nine WES-undiagnosed children, 100% of potentially pathogenic loci in their candidate gene list were covered (hence no bar). The variability in the fractions falling below the three quality thresholds is reflective of the different genic content of individual gene lists. Potentially pathogenic loci are defined as those with any entry with attributed pathogenicity in HGMD or ClinVar at the time of analysis [28, 29]. A threshold of 10× represents the lower limit of variant curation for our study, 20× represents a commonly cited threshold at which 99% of single-nucleotide variants (SNVs) are detected [–32]