Exome and Whole-Genome Sequencing as Clinical Tests: A Transformative Practice in Molecular Diagnostics

In the last few years, next-generation sequencing (NGS)5 has revolutionized the approaches by which we interrogate the genetic causes of rare single-gene disorders (1). More recently, NGS has been rapidly moving into the clinical diagnostics arena and transforming the practice of molecular diagnostics testing. Many NGS-based tests for multigene panels are available in clinical laboratories. Now, proof-of-principle studies are demonstrating the feasibility of using exome sequencing for clinical testing (2–4). Dixon-Salazar et al. used exome sequencing to interrogate a cohort of 118 probands with recessive neurodevelopmental disorders and identified disease-causing mutations in 19% of the probands. Importantly, exome sequencing uncovered mutations in 10 probands (8% of the cohort) that altered the initial clinical diagnosis and led in some cases to a change in patient management (2). Calvo et al. tested 42 infants with mitochondrial diseases with the “MitoExome” (mitochondrial DNA plus exons of approximately 1000 nuclear genes) and confirmed that 10 patients (24% of the cohort) had pathogenic mutations. In addition, these investigators identified 2 novel disease-causing genes (3). Need et al. used exome sequencing in family trios to investigate patients with a broad range of phenotypic presentations and reported a 33% diagnostic yield: 4 of 12 patients with apparently causal mutations and 2 additional potentially novel disease-causing genes (4). Although the number of patients tested thus far by whole-exome sequencing (WES) or whole-genome sequencing (WGS) is still small, the data reported in these recent studies have defined the minimal diagnostic yields of such NGS-based clinical tests at their early stage of implementation, thus demonstrating the practicality of NGS-based tests. We anticipate that the clinical utility of NGS-based tests will improve further in the next few years and that more patients will benefit from improved diagnostic accuracy and more-appropriate treatment. Conventional molecular diagnostics tests evaluate a defined number of exons for one or more genes, the selection of which is based on the best clinical judgment after thorough evaluation of the patient's clinical presentation. These tests, which are performed in CLIA-certified laboratories, are mainly for mendelian disorders with well-established causal relationships between genes and diseases. NGS-based comprehensive gene panels, WES-based tests, or WGS-based tests, however, assay for a substantially larger number of genes and genomic regions, only a small proportion of which would potentially be associated with the patient's condition. NGS-based tests are performed mainly for patients without a clear clinical diagnosis or for patients who have negative test results for genes known to be associated with the disorder. Clinical diagnostics laboratories are actively engaged in implementing NGS-based tests. We review some important characteristics of NGS-based tests and discuss a few outstanding analytical issues, with the hope of gaining a better understanding of the nature of this transformative diagnostic approach and achieving a glimpse into the future of NGS-based tests. The sequences targeted with commercial exome-capture reagents differ from one vendor design to another. For example, the total number of bases targeted by SureSelect V4 (Agilent Technologies) is 51 Mb (71 Mb for SureSelect V4+UTRs) comprising 20 965 genes and 334 378 exons, whereas the number of bases is 62 Mb with Illumina TruSeq (20 794 genes and 201 121 exons) and 64 Mb with NimbleGen's SeqCap EZ Human Exome Library v3.0 (>20 000 genes). A portion of the exome (such as exon 1 and 5′ untranslated regions) is known to be variably resistant to different capturing methods. Even when the same exome-capturing kit is used, capture efficiency differs from one experiment to another, as does the total number of sequence reads generated by each run. In addition, different variant-calling algorithms—such as BWA, novoalign, and BWT—will generate largely overlapping but substantially different variant sets. Consequently, different numbers of variants will be detected for the same sample when different exome-capturing kits are used or when a sample is run in different laboratories, or even at different times. Such variation is a major deviation from the traditional performance characteristics of molecular testing, in which minimal interlaboratory and interrun differences and 100% reproducibility are expected. The unique performance characteristics of WES/WGS have important implications for how we validate an NGS-based clinical test and how to perform and regulate quality control. Validation for NGS-based tests proves to be difficult for a variety of reasons. For example, NGS technologies are still evolving rapidly, new methods are constantly being developed, and existing methods are often replaced with newer versions of protocols before they can be thoroughly validated. The high costs and huge workloads associated with validation by traditional approaches (i.e., Sanger sequencing) preclude comprehensive evaluations of analytical sensitivity, specificity, and reproducibility. We expect this situation to improve as the assays become more standardized and consistent in the future, but a modified validation protocol should be used for developing an NGS-based test. In contrast to conventional tests, which mostly target only known disease genes, WES/WGS tests assay all genes or the whole genome. Intrinsically, WES/WGS is a process that presents both discovery and diagnostic functionalities. In many cases, discoveries are directly translated into diagnostics when compelling evidence exists for establishing causal relationships between novel variants and phenotypes. Currently, such supporting evidence includes detecting a recurring de novo mutation in a gene with a deleterious effect on the encoded protein in similarly affected patients, showing cosegregation in family members, and/or providing compelling evidence of the underlying biology and mechanism. The dual functionality of WES/WGS has direct implications, not only for the selection of patients and family members but also for data analysis. The current data-analysis process typically involves filtering all of the variants (generally approximately 24 × 103 to 28 × 103) by means of the following steps (2): (a) Variants not occurring in exons or at potential splicing sites are removed (currently, even WGS data are initially filtered this way, effectively eliminating the major added value of doing WGS in the first place) (5). This step leaves approximately 16 × 103 to 21 × 103 variants. (b) Variants that are not nonsynonymous, do not occur at splicing sites, and do not produce insertions/deletions are removed. This step leaves approximately 8 × 103 to 12 × 103 variants. (c) Common variants that occur in public databases, such as dbSNP, are removed. (d) The remaining variants (approximately 200 to 600) are then compared with an in-house database—as in the study of Dixon-Salazar et al. (2)—to narrow the variants of interest further to approximately 100 to 200 variants. From this point onward, the inheritance pattern, information on segregation, a variant's functional annotation, and correlation with a predicted disease mechanism are required to implicate the most likely candidate gene. For dominant disorders, parental samples are very helpful for identifying de novo variants. For both dominant and recessive conditions, additional family members can help narrow down the number of candidate genes substantially. Given the data-analysis pipeline described above, further pinpointing of the causal variant(s) from a small list of candidate genes proves to be crucial. We anticipate that gene expression profiling with high-throughput whole-transcriptome sequencing technologies (i.e., RNA-seq) and high-throughput functional assays could eventually be part of an extended clinical diagnostics process to help with this critical step. Unlike the “targeted array vs whole-genome array” debate, which was largely a philosophical one, the debate regarding the choice of WES or WGS is largely focused on pragmatic issues. Seven years ago, the clinical diagnostics community debated whether we should be detecting and reporting copy number variants for which we had no knowledge of their clinical importance. Today, emboldened by the success of the clinical utility of whole-genome microarray testing, the community no longer debates this question of principle. The debate of WES vs WGS mainly revolves around the price tag and related issues, such as turnaround time, the labor involved in data generation, and data storage requirements. As the price of generating whole-exome and whole-genome data continues to drop and the price difference between WES and WGS continues to narrow, WGS is anticipated to become more widely used in the clinical setting. Currently, however, gene panel–based tests and WES are leading the way in clinical diagnostics. The full potential of WGS can be realized only when we gain a much better understanding of the functions of noncoding regions. In our opinion, WES, WGS, and comprehensive gene panel tests should remain as options to be considered for different disease groups and be used according to a patient's specific conditions. Alternatively, a combined approach can be used to capture all variety of genomic variations. That is, an upgraded exome that includes all coding regions and all noncoding regulatory sequences will be used to detect single-nucleotide polymorphisms, insertions/deletions, and large intragenic copy number variants, and a low-cost (i.e., low-coverage) WGS will be used to detect potential structural rearrangements. In summary, although NGS-based molecular diagnostics tests are still in their infancy, they have demonstrated excellent clinical utility for single-gene disorders. With further developments in NGS technologies for data generation and with more effective bioinformatics tools for data analysis and clinical extraction, the full potential of WES/WGS that we expect to be revealed in the coming years will greatly enrich and empower the practice of genomic medicine beyond the rare single-gene disorders. The improvements in patient care demonstrated in recent studies justify all the effort and cost for moving these new and exciting approaches into molecular diagnostics practice. next-generation sequencing whole-exome sequencing whole-genome sequencing. The authors thank Drs. J.F. Gusella and O. Platt for revising the manuscript and for their comments.