Lessons learned: next-generation sequencing applied to undiagnosed genetic diseases

Rare genetic disorders, when considered together, are relatively common. Despite advancements in genetics and genomics technologies as well as increased understanding of genomic function and dysfunction, many genetic diseases continue to be difficult to diagnose. The goal of this Review is to increase the familiarity of genetic testing strategies for non-genetics providers. As genetic testing is increasingly used in primary care, many subspecialty clinics, and various inpatient settings, it is important that non-genetics providers have a fundamental understanding of the strengths and weaknesses of various genetic testing strategies as well as develop an ability to interpret genetic testing results. We provide background on commonly used genetic testing approaches, give examples of phenotypes in which the various genetic testing approaches are used, describe types of genetic and genomic variations, cover challenges in variant identification, provide examples in which next-generation sequencing (NGS) failed to uncover the variant responsible for a disease, and discuss opportunities for continued improvement in the application of NGS clinically. As genetic testing becomes increasingly a part of all areas of medicine, familiarity with genetic testing approaches and result interpretation is vital to decrease the burden of undiagnosed disease.

Genetic testing approaches vary based on clinical indications and have evolved over time to include NGS. In 1999, the Task Force on Genetic Testing defined a genetic test as “the analysis of human DNA, RNA, chromosomes, proteins, and certain metabolites in order to detect heritable disease-related genotypes, mutations, phenotypes or karyotypes for clinical purposes. Such purposes include predicting risk of disease, identifying carriers and establishing prenatal and clinical diagnosis or prognosis. Prenatal, newborn and carrier screening, as well as testing in high-risk families, are included” (11).

Newborn screening. In many countries, newborn screening (NBS) is used to rapidly identify neonates with treatable genetic conditions. NBS uses a combination of tandem mass spectrometry, gel electrophoresis, enzymatic activity assays, and gene sequencing (12). The goal of NBS is to rapidly identify neonates at high risk for prompt evaluation and treatment. Inclusion of disorders in NBS varies based on location, but typically includes disorders of amino acid metabolism (e.g., phenylketonuria, maple syrup urine disease), disorders of fatty acid oxidation (e.g., medium-chain acyl-CoA dehydrogenase deficiency), disorders of carbohydrate metabolism (e.g., galactosemia), hemoglobinopathies (e.g., sickle cell anemia), and cystic fibrosis. Importantly, NBS is a screening test, not a diagnostic test. It requires rapid assays of optimally collected blood spot samples and prompt follow-up testing to solidify a diagnosis.

Biochemical studies. Laboratory biochemical evaluations are used to assess for inborn errors in metabolism (IEM) because of their greater sensitivity (in some cases) and faster turnaround time compared with NGS. They include tests such as acylcarnitine profiles, plasma amino acids, urine organic acids, ammonia levels, free and total carnitine, creatine metabolites, vitamin levels, and complete metabolic profiles. They can provide biochemical evidence for an underlying IEM and, in some cases, be diagnostic without a confirmatory molecular test. As there is overlap in the techniques used by NBS and these biochemical studies, there is overlap in detectable diagnoses. Biochemical studies can be used for confirmation of NBS results and/or uncover disorders of creatine metabolism, urea cycle disorders, carnitine deficiency, and mitochondrial dysfunction. A study showed that, while NGS was able to diagnose 50% of the cases with a suspected IEM, specific biochemical profiles can provide phenotypic data that increase the likelihood of a diagnosis with NGS (13).

Karyotype and chromosomal microarray. For decades, karyotypes were the only means to identify chromosomal abnormalities including aneuploidies (abnormal numbers of chromosomes) and large (more than 3–10 Mb) deletions or duplications (Table 1 and refs. 14, 15). With the development of chromosomal microarrays (CMAs), there has been a large shift from the use of karyotypes to the use of CMAs as a first-line genetic test in evaluating individuals with developmental delay, intellectual disability, multiple congenital anomalies, and autism spectrum disorders (15–17). While there are several different types of CMAs, all use fluorescence to genotype and count the number of alleles at millions of locations across the genome. As a product of this resolution, CMAs have increased sensitivity for deletions and duplications, with newer platforms detecting copy number variants (CNVs) as small as 30 kb (Table 1, Figure 1, and ref. 15). CMAs will still fail to detect balanced chromosomal translocations and have poor sensitivity in detecting inversions and triploidy; karyotypes are still better for the detection of translocations (Figure 1) and aneuploidies. While the ability of sequencing analysis to detect CNVs continues to improve, short-read, exon-only sequencing platforms still miss many CNVs (Table 1 and ref. 18).

Figure 1

Schematic diagrams representing different types of genomic variation. By column from left to right are diagrams depicting a gene schematic with a variant, the type of variant, and a specific example of that type of variant. Those columns are followed by example genes that have been demonstrated to harbor that kind of variant, a representative disease caused by that type of variant, and the OMIM reference for that disease. The last column additionally includes references to cases in this Review, if applicable.

Table 1

Ability of genetic testing approaches to detect various types of genomic variation

Sequencing. The practice of clinical genetics has evolved in parallel with testing strategies and knowledge of the genome. Historically, many diagnoses were based on clinical diagnostic criteria rather than a molecular test. As more was learned about phenotypic and genotypic diversity of genetic disorders and our ability to sequence more genes in a cost- and time-efficient manner evolved, practice shifted from the use of single-gene sequencing to gene panels in which multiple genes whose dysfunction could cause overlapping phenotypes are sequenced. Eventually these gene panels incorporated NGS for disorders that were not detected by karyotype, CMA, or fluorescence in situ hybridization (FISH) analyses (Table 1). Single-gene sequencing or gene panels are often used when specific diagnoses are suspected (e.g., neurofibromatosis, Noonan syndrome, or CHARGE syndrome) or when there are phenotypes that overlap between genetic disorders (e.g., congenital heart disease, autism spectrum disorders, or skeletal dysplasias).

Sequencing tests can identify variants that are defined as nucleotide differences from the reference sequence. Missense variants are those that produce a single amino acid change (Figure 1). Nonsense variants are those that result in premature stop codons leading to early termination of the protein sequence (Figure 1). Subsequently truncated proteins can either be dysfunctional or be subjected to nonsense-mediated decay. Splicing variants interfere with pre-mRNA processing so that the final mRNA that is translated includes inappropriate intronic sequences and/or excludes exonic sequences (Figure 1). Promoter variants can alter the affinity of RNA polymerase for the promoter site, which can affect the amount of mRNA produced. Synonymous variants change the DNA sequence at the codon level, but the amino acid encoded by the new codon does not change because of redundancy in the genetic code. The ability to detect these single-nucleotide variants differs between sequencing platforms. Exon-based sequencing should detect missense, nonsense, and synonymous variants and may or may not detect variants that affect splicing, depending on the location of the variant and how much of the intron/exon boundary is captured. ES approaches tend to miss regulatory or promoter variants. GS should capture all types of single-nucleotide variants (Table 1).

In addition to determining the effect the variant has on the expression and/or processing of the encoded mRNA and its protein product, the clinical impact of a variant can be difficult to determine; this is especially true for missense and synonymous variants. The American College of Medical Genetics and Genomics (ACMG) has provided guidelines for variant classification that weighs evidence to classify variants as pathogenic, likely pathogenic, of uncertain significance, likely benign, or benign (19).

Evidence for variant classification is gathered from published literature, bioinformatics databases, and in silico tools. Population databases, like the Genome Aggregation Database (gnomAD) (20), report variant frequencies in the general population to address the thought that pathogenic variants should appear at a population frequency corresponding to the prevalence of the genetic disease. GnomAD has data from more than 125,000 exomes and more than 15,000 genomes from individuals who are not biologically related, represent a variety of ancestries, and are presumably healthy (20). Conversely, disease databases, like ClinVar, contain variants found in patients with genetic disease. ClinVar has information on more than 500,000 variants, including phenotypic data and clinical interpretation (21, 22). Phenotypic information, whether reported in primary literature or databases, is crucial for genotype-phenotype correlation and thus variant interpretation. Finally, in silico models are used to predict the effects of a variant at the nucleotide, splicing, or amino acid level. The many different informatics tools available have varying strengths and weaknesses. These tools make predictions on pathogenicity based on evolutionary conservation at the amino acid level, deviation from known splicing motifs, effects on amino acid sequence, or how dissimilar the properties of the variant amino acid are versus the reference amino acid (20). Each of these sources of evidence has its shortcomings. For instance, there are individuals thought to be healthy who harbor pathogenic variants; databases are not necessarily peer-reviewed or updated frequently; and in silico prediction tools do not fully capture biological complexity and therefore may be inaccurate. Despite use of multiple tools, there is often insufficient evidence to determine whether a variant is disease-causing at the time of its discovery.

As use of ES has increased, it has been demonstrated to be superior to gene panels in diagnostic rate and, in some cases, cost (23). The overall reported diagnostic yield of ES varies based on institution, when the analysis was done, and the clinical indication for testing. The diagnostic rate has continued to increase from 25% in 2013–2014 (6, 7) to 28.8%–31.0% in 2016 (5). A study in 2017 demonstrated a diagnostic rate of 52% using singleton ES in children with multiple congenital anomalies (24). The continuing evolution in the use of NGS was demonstrated when GS was shown to have a diagnostic rate of 41% compared with 24% with other genetic testing approaches, including ES. The authors concluded that the ability of GS to identify structural variants and noncoding variants (NCVs) that were not detected by ES platforms explained this difference (Table 1 and ref. 18).

Sometimes a patient’s phenotype is caused by a combination of genetic diseases. Such presentations can confound NGS data interpretation, because the initial impulse is to assign one molecular cause to all the patient’s phenotypes. This can lead to diagnostic delays or inaccurate attribution of phenotypes to the wrong candidate variant and subsequent assignment of incorrect diagnoses to other patients with similar symptoms (25). One study found that 4.9% of patients who underwent ES had multiple molecular diagnoses (26). Correct phenotyping is key to NGS variant analyses and is vital to the detection of candidate variants (27–30).

While there are clinical scenarios in which NGS improves the diagnostic rate, there are others in which alternative approaches are still important and/or the standard of care. In the following section, we present several examples that illustrate different pitfalls that prevented NGS alone from providing a diagnosis for patients (summarized in Table 2). The following previously published examples are cases from our experiences and other similar experiences in which NGS was not the primary means of making a genetic diagnosis.

Table 2

Summary of the lessons learned from cases in which NGS was not sufficient to identify the genetic change causing the genetic disease

Case 1: detecting copy number variants. A 60-year-old woman with multiple benign neck paragangliomas, episodic hypertension with tachycardia, vocal cord paralysis, and an extensive family history of paragangliomas was tested with a hereditary pheochromocytoma and paraganglioma (HPP) NGS panel, which was negative. She and two of her family members then had GS, which detected a 2.17 kb deletion that included exon 5 of the SDHD gene (Table 2 and Figure 1). Subsequent high-density CMA confirmed the deletion, resulting in a diagnosis of paragangliomas 1 (31) (OMIM #168000; ref. 32). Deletions in SDHD account for as many as 10% of HPP-causing variants (33). While SDHD was included in the original panel, the testing did not include deletion and duplication (i.e., copy number variant) testing. NGS testing, including gene panels, ES, and GS, can detect some copy number variants (CNVs), but dedicated CNV testing should be considered if it was not included with the single-gene or gene-panel sequencing or if the sequencing test failed to uncover a CNV and the clinical suspicion remains high (34).

Case 2: detecting mosaicism. A 5-year-old boy with developmental delay, atrial septal defect, hypotonia, and skin pigmentation variation was found to have diploid/triploid mosaicism (DTM) via GS of skin biopsies. DTM occurs when some cells have three sets of chromosomes (triploid) and the remainder have the normal two copies (diploid). The patient’s GS diagnosis was confirmed by a karyotype (31). While both a karyotype and GS resulted in the same diagnosis, testing could have started with a CMA for his developmental delay based on 2010 ACMG guidelines (35). A CMA would have identified this patient’s mosaicism, which could then be confirmed by a karyotype. Retrospectively, this testing strategy could have prevented GS, which is a relatively time-consuming and expensive test in comparison with CMA or karyotype (Tables 1 and 2). It could alternatively be argued that the patient’s variation in skin pigmentation could be a manifestation of underlying mosaicism and that an NGS strategy as first-tier testing on a limited source of DNA (i.e., the skin biopsy) would be indicated, especially in the context of the recent recommendations by the ACMG (36). The ideal testing strategy is not always obvious based on patient presentation, but continued consideration of what various genetic tests can and cannot detect can increase the likelihood of a diagnosis. In cases in which the chromosomal change is too small to be identified with a karyotype, CMA or NGS would be needed to identify mosaicism (31).

Case 3: detecting noncoding variants; using transcript analysis. A 5-year-old girl was seen for developmental delay starting at 3 months of age. She had inversion of her feet that progressed to muscle weakness, calf atrophy, and decreased lower extremity reflexes at 18 months of age. She had normal basic and metabolic biochemical laboratory evaluations, electromyography and nerve conduction studies that were consistent with motor axonal polyneuropathy, and normal MRIs of her brain and spine. She had genetic testing for spinal muscular atrophy type 3 that was negative. ES revealed a single pathogenic variant in IGHMBP2, a gene associated with an autosomal recessive type of Charcot-Marie-Tooth disease (OMIM #616155; ref. 32). No other variants were reported. GS revealed a deep-intronic noncoding variant (NCV), also in IGHMBP2 (Figure 1). Reverse transcriptase PCR revealed that this variant activated a cryptic splice site and led to a frameshift insertion that resulted in a premature termination codon, and that nonsense-mediated decay caused destruction of the IGHMBP2 transcripts. It was necessary in this case to use GS and transcript analysis to identify and prove the functional impact of the IGHMBP2 NCV (Tables 1 and 2 and ref. 37).

Case 4: FISH and cosegregation studies detect translocations. A 3-year-old girl with a history of multiple café au lait macules and cervicomedullary and retropharyngeal plexiform neurofibromas met the clinical criteria for diagnosis of neurofibromatosis type 1 (NF-1) (OMIM #162200) (32). When she was a neonate, karyotype revealed a balanced translocation between her chromosomes 4 and 17 (Figure 1). Because the NF1 gene is on chromosome 17, NGS assessed for sequencing variants in NF1 and multiplex ligation-dependent probe amplification (MLPA) was used to look for CNVs; both tests were negative. The patient’s mother and two brothers also had clinical diagnoses of NF-1 and shared the balanced translocation. Since NF-1 cosegregated with the balanced translocation in this family, it was suspected that the translocation breakpoint disrupted the NF1 allele (Table 2 and Figure 1). This was confirmed using FISH analysis with custom probes for both the 5′ and 3′ regions flanking the NF1 gene (38). Gene translocations can be missed by NGS and MLPA (Table 1). With the incorporation of multiple testing approaches including NGS, the molecular diagnostic rate for NF-1 has increased from approximately 50% to approximately 95% (38).

Case 5: detecting methylation and imprinting variants. A 12-year-old girl was evaluated for progressive obesity, hypotonia, recurrent fractures, and developmental delays. Genetic testing included CMA, Prader-Willi methylation, mucopolysaccharidosis biochemical testing, and GNAS and WFS1 sequencing. ES was non-diagnostic but detected a pathogenic ANO5 variant and variants of uncertain significance in ELN, HIVEP2, COL6A3, and LRP5. ANO5 was the top candidate of interest because of the patient’s history of pathological fractures and muscle hypotonia. A second pathogenic variant, which would be expected in an autosomal recessive disorder, was not detected despite deletion/duplication studies of ANO5. More detailed phenotyping and a literature review suggested that she might have phenotypic manifestations of Temple syndrome (a disorder caused by abnormal methylation). Methylation analysis at the MEG3 transcriptional start site differentially methylated region on chromosome 14q32 showed complete hypomethylation (Figure 1), which was both different from the normal heterozygous methylation pattern and consistent with a diagnosis of Temple syndrome (31) (OMIM #616222; ref. 32). ES and GS will miss epigenetic disorders, and careful phenotyping combined with a high index of clinical suspicion is required to make these diagnoses (Tables 1 and 2). Therefore, methylation studies should be considered concurrently with NGS when warranted. This case illustrates why genetic evaluations should be iterative processes that include reevaluation of the phenotypes and genotypes that are used to generate differential diagnoses.

Case 6: detecting repeat expansions. A kindred was identified with a strong family history of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). There were ten affected individuals in an autosomal dominant inheritance pattern. These individuals experienced symptoms including loss of empathy and apathy, impaired executive function, dysphagia, apraxia, and hyperreflexia. Neuropathological findings revealed atrophy of the frontal lobes with variable involvement of other regions, neurodegeneration, and decreased myelination. Linkage analysis implicated a region on chromosome 9p as responsible for conferring this phenotype. Despite there being only 10 genes in this region, genomic sequencing and expression analysis failed to identify the gene responsible for the ALS-FTD phenotypes (39). Subsequent studies identified that a hexanucleotide repeat in C9orf72 was responsible for ALS-FTD (40, 41) (OMIM #105550; ref. 32). A similar case of FTD was reported in which both ES and GS were non-diagnostic, but repeat expansion (RE) testing of C9orf72 revealed more than 44 repeats, which led to the diagnosis (31). Recent work has improved the ability to identify REs from NGS data. Initial bioinformatics approaches could detect REs from NGS data if the location of the RE of interest was already known (42). Subsequent work improves detection of REs in a way that does not rely on knowing that an RE exists by leveraging the unique sequence flanking the RE to sequence into the RE. However, these approaches that use short-read sequencing do not give insight into the length or composition of the RE (43). Leveraging multiple RE detection approaches simultaneously has improved RE detection, with some caveats. GS more reliably detects REs than does ES. A combination of genotyping and statistical approaches can increase the reliability of RE detection. Reliable determination of RE length and composition remains a problem, and current RE detection often still relies on PCR or Southern blot techniques (Tables 1 and 2, Figure 1, and ref. 44).

We have discussed types of genomic changes missed with NGS and have provided case examples in which other approaches were required to make a diagnosis (Figure 1 and Table 1). Further advancements in NGS technologies will continue to improve diagnostic utility. For example, the ACMG released a practice guideline in 2021 recommending that ES/GS be considered a first- or second-tier test in patients with congenital anomalies, developmental delay, or intellectual disability based on clinical utility for providers and families (35). Prior to this guideline, there was variability in the diagnostic approach for these patients; first-tier testing could include CMA, fragile X testing, and biochemical studies. Single-gene or gene-panel testing could be incorporated at any point in the diagnostic workup. The new guidelines are based on evidence demonstrating utility in performing ES or GS after CMA or focused testing. Even so, testing options are influenced by medical insurance, provider preferences, parental desires, and health care system policies and practices (35).

Short-read versus long-read sequencing. DNA sequencing technology has changed greatly in the more than 40 years since Sanger sequencing was developed. Over this time, sequencing has transitioned from gel electrophoresis–based approaches like Sanger sequencing to shotgun sequencing to the NGS approach that uses massively parallel sequencing (45). NGS has greatly decreased the cost and increased the efficiency of DNA sequencing. These technologies are primarily “sequence by synthesis,” where complementary nucleotides are added sequentially, causing nucleotide-specific fluorescence that is read by a camera and results in sequences that are 100 to 200 bp in length (46). While this approach has revolutionized the fields of genetics and genomics, short reads can be difficult to computationally align to the reference genome, which makes resolution of complex and repetitive regions of genomes difficult and severely limits detection of structural variants (Table 1, Figure 1, and refs. 47, 48). In contrast, evolving long-read sequencing approaches generally use alternative sequence by synthesis chemistry or measurement of changes in electrical current caused by the DNA molecule as it passes through a nanopore. These reads can be 10 kb to several Mb in length (46). Further advances to increase the accuracy and decrease the cost of long-read sequencing technologies could lead to a genetic testing approach that would allow for de novo genome assembly; identify sequence variation, copy number variation, REs, and structural variants; and provide more accurate phasing of variants (Table 1, Figure 1, and refs. 46, 49, 50–53). However, obstacles to realizing the full potential of long-read sequencing include that it remains relatively expensive (6- to 12-fold more expensive by some estimates) and does not have the base-calling accuracy of short-read sequencing technology (50, 53). Data storage and scalability are also issues; large genomes present the problem of storing large amounts of data as well as demonstrate decreasing efficiency of genome assembly as genome size increases. While long-read sequencing can identify epigenetic changes, these technologies have not yet realized that full potential (54). Short-read sequencing covers a majority of the known, disease-causing structural variants, and consequently it is thought that the addition of long-read sequencing technology, in its current state, is unlikely to substantially increase diagnostic yield (50). As the sequencing technology continues to improve, the cost decreases, and the bioinformatics pipelines become more accurate and efficient, long-read sequencing may eventually contribute to an increased diagnostic yield from genetic testing (46, 50, 53, 54).

Data sharing and cloud computing. Another opportunity to improve our understanding of genomic function and dysfunction comes from leveraging large genomic databases that are coupled with phenotypic data. Currently, ClinVar and gnomAD are two of these most-used databases. ClinVar has partnered with a collaborative program called Clinical Genome Reference (ClinGen) to improve the curation, sharing, and archiving of genomic variation data as well as their clinical interpretation or relevance. ClinGen curates data for ClinVar from other databases and structures data submissions into proper format and nomenclature. The program also developed a system to define the review level of submissions. To aid in reviewing submissions, ClinGen develops expert teams in various clinical realms to validate variant pathogenicity and gene-disease relationships (55).

An example of the power of using large databases comes from Brokamp et al., who reported a patient in whom they identified a de novo frameshift variant that had not been observed previously (56). There were no matches in the available matching tools (GeneMatcher, MyGene2, Matchmaker Exchange; refs. 57–59) or in ClinVar and gnomAD. However, by utilizing their in-house database of more than 3 million individuals’ electronic health records, many of which had accompanying genomic data (i.e., BioVU; ref. 60), they found two other individuals with de novo variants in the same gene and overlapping phenotypes. By identifying multiple, unrelated individuals with variants in the same gene and with very similar phenotypes, they were able to change the designation of the variant from one of uncertain significance to pathogenic and discover a new genetic disorder (56).

The case above illustrates the potential of leveraging large data to identify other exceptionally rare cases to make diagnoses. NGS has rapidly increased the amount of available genomic data, which presents both opportunities and difficulties associated with working with petabytes (1 petabyte = 1 million gigabytes) of data to solve cases. Unfortunately, the infrastructure necessary to utilize a data set of this size is prohibitive to most clinicians and independent laboratories. Cloud computing is a system in which resources are rented to mitigate the need to establish both the hardware and software necessary for data analysis of this magnitude (61). Addressing the data sharing protocols and patient privacy concerns that come with cloud computing will be necessary to be able to utilize these platforms to their full potential.

The All of Us Research Program is an example of using cloud computing to facilitate the application of genomics in health care. This program plans to provide a resource of genomic and phenotypic data of at least 1 million people, most of whom are from backgrounds underrepresented in biomedical research. The goal of the All of Us Research Program is to create a resource of health questionnaires, electronic health record data, physical measurements, and both digital data and biospecimens for a variety of applications including characterizing natural histories of diseases, identifying disease risk factors, and revealing new biomarkers. The design of the program should mitigate the small sample sizes and lack of diversity in many genomic data sets that limit medical discovery (62). While the intention is not directly for the diagnosis of rare and/or undiagnosed disease, study participants will have the option of learning about pharmacogenomic findings as well as actionable, highly penetrant, disease-causing variants (62).

Sequencing critically ill pediatrics patients. Much of this Review has discussed the application of genetic testing to improve the diagnostic rate in patients for whom there is a concern regarding an underlying genetic disorder. However, critically ill patients, who may not yet present the classic signs or symptoms of a rare and unfamiliar genetic disorder, present another opportunity for the application of NGS to detect an undiagnosed genetic disease. Many severe genetic conditions present in the neonatal period or in early childhood, but the onset of characteristic signs and symptoms is delayed because they are age dependent. Multiple studies of the utility of NGS in critically ill neonates have shown increased diagnostic rates, decreased costs associated with hospitalization, changes in management, and increased patient and family satisfaction. Studies that obtained NGS on critically ill pediatric patients with concerns regarding an underlying genetic disorder yielded diagnostic rates of 21% to 58% depending on patient selection, year the study was conducted, and NGS methodology (63–72). These studies defined clinical utility as changes in medical or surgical management, testing family members for related genotypes or phenotypes, informing recurrence risk, suggesting a potential pharmaceutical, and/or involving palliative care. They reported that, in 21%–83% of cases, NGS led to a change in management regardless of whether a diagnosis was made (63–68, 70–72). For those patients in whom there was a suspicion of a genetic disease, studies showed that either ES or GS yielded an increased diagnostic rate when compared with gene panel alone or standard genetic testing approaches: 58% with trio-based ES versus 12.5%–25% with gene panels (64), 57% with ES versus 13.75% with standard approaches (66), and 57% with GS versus 9% with standard approaches (70). The Newborn Sequencing in Genomic Medicine and Public Health randomized controlled trial 1 (NSIGHT1) was a program that tested the hypothesis that rapid GS “increased the proportion of [critically ill] infants receiving a genetic diagnosis within 28 days.” NSIGHT1 was terminated early because GS demonstrated an obvious clinical benefit compared with the standard approaches (73). Finally, one study showed that, in patients with a low suspicion of an underlying genetic disorder, NGS achieved a genetic diagnosis in 53% of their cases (71).

This work was supported by NIH grants 1U01HG007674 and R38-HL143619.

Address correspondence to: Bryce A. Schuler or Rizwan Hamid, Division of Medical Genetics & Genomic Medicine, 161 21st Ave South, DD-2205 Medical Center North, Nashville, Tennessee 37232-2578, USA. Email: Bryce.a.schuler@vumc.org (BAS). Email: Rizwan.hamid@vumc.org (RH).

Conflict of interest: The authors have declared that no conflict of interest exists.

Copyright: © 2022, Schuler et al. This is an open access article published under the terms of the Creative Commons Attribution 4.0 International License.

Reference information: J Clin Invest. 2022;132(7):e154942.https://doi.org/10.1172/JCI154942.

1. Nguengang Wakap S, et al. Estimating cumulative point prevalence of rare diseases: analysis of the Orphanet database. Eur J Hum Genet. 2020;28(2):165–173.
   View this article via: CrossRef PubMed Google Scholar
2. Clark MM, et al. Meta-analysis of the diagnostic and clinical utility of genome and exome sequencing and chromosomal microarray in children with suspected genetic diseases. NPJ Genom Med. 2018;3:16.
   View this article via: CrossRef PubMed Google Scholar
3. Farwell KD, et al. Enhanced utility of family-centered diagnostic exome sequencing with inheritance model-based analysis: results from 500 unselected families with undiagnosed genetic conditions. Genet Med. 2015;17(7):578–586.
   View this article via: CrossRef PubMed Google Scholar
4. Lee H, et al. Clinical exome sequencing for genetic identification of rare Mendelian disorders. JAMA. 2014;312(18):1880–1887.
   View this article via: CrossRef PubMed Google Scholar
5. Retterer K, et al. Clinical application of whole-exome sequencing across clinical indications. Genet Med. 2016;18(7):696–704.
   View this article via: CrossRef PubMed Google Scholar
6. Yang Y, et al. Clinical whole-exome sequencing for the diagnosis of mendelian disorders. N Engl J Med. 2013;369(16):1502–1511.
   View this article via: CrossRef PubMed Google Scholar
7. Yang Y, et al. Molecular findings among patients referred for clinical whole-exome sequencing. JAMA. 2014;312(18):1870–1879.
   View this article via: CrossRef PubMed Google Scholar
8. Bacchelli C, Williams HJ. Opportunities and technical challenges in next-generation sequencing for diagnosis of rare pediatric diseases. Expert Rev Mol Diagn. 2016;16(10):1073–1082.
   View this article via: CrossRef PubMed Google Scholar
9. Pereira R, et al. Bioinformatics and computational tools for next-generation sequencing analysis in clinical genetics. J Clin Med. 2020;9(1):E132.
   View this article via: PubMed Google Scholar
10. Liu Z, et al. Toward clinical implementation of next-generation sequencing-based genetic testing in rare diseases: where are we? Trends Genet. 2019;35(11):852–867.
    View this article via: CrossRef PubMed Google Scholar
11. Holtzman NA, Watson MS. Promoting safe and effective genetic testing in the United States. Final report of the Task Force on Genetic Testing. J Child Fam Nurs. 1999;2(5):388–390.
    View this article via: PubMed Google Scholar
12. Watson MS, et al. Newborn screening: toward a uniform screening panel and system—executive summary. Pediatrics. 2006;117(5 pt 2):S296–S307.
    View this article via: PubMed Google Scholar
13. Ghosh A, et al. Diagnosing childhood-onset inborn errors of metabolism by next-generation sequencing. Arch Dis Child. 2017;102(11):1019–1029.
    View this article via: CrossRef PubMed Google Scholar
14. Bi W, et al. Comparison of chromosome analysis and chromosomal microarray analysis: what is the value of chromosome analysis in today’s genomic array era? Genet Med. 2013;15(6):450–457.
    View this article via: CrossRef PubMed Google Scholar
15. Miller DT, et al. Consensus statement: chromosomal microarray is a first-tier clinical diagnostic test for individuals with developmental disabilities or congenital anomalies. Am J Hum Genet. 2010;86(5):749–764.
    View this article via: CrossRef PubMed Google Scholar
16. Gonzales PR, et al. Overview of clinical cytogenetics. Curr Protoc Hum Genet. 2016;89:8.1.1–8.1.13.
    View this article via: CrossRef PubMed Google Scholar
17. Wapner RJ, et al. Chromosomal microarray versus karyotyping for prenatal diagnosis. N Engl J Med. 2012;367(23):2175–2184.
    View this article via: CrossRef PubMed Google Scholar
18. Lionel AC, et al. Improved diagnostic yield compared with targeted gene sequencing panels suggests a role for whole-genome sequencing as a first-tier genetic test. Genet Med. 2018;20(4):435–443.
    View this article via: CrossRef PubMed Google Scholar
19. Richards S, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015;17(5):405–424.
    View this article via: CrossRef PubMed Google Scholar
20. Koch L. Exploring human genomic diversity with gnomAD. Nat Rev Genet. 2020;21(8):448.
    View this article via: CrossRef PubMed Google Scholar
21. Biesecker LG, et al. The ACMG/AMP reputable source criteria for the interpretation of sequence variants. Genet Med. 2018;20(12):1687–1688.
    View this article via: CrossRef PubMed Google Scholar
22. Landrum MJ, et al. ClinVar: improvements to accessing data. Nucleic Acids Res. 2020;48(d1):D835–D844.
    View this article via: CrossRef PubMed Google Scholar
23. Dillon OJ, et al. Exome sequencing has higher diagnostic yield compared to simulated disease-specific panels in children with suspected monogenic disorders. Eur J Hum Genet. 2018;26(5):644–651.
    View this article via: CrossRef PubMed Google Scholar
24. Tan TY, et al. Diagnostic impact and cost-effectiveness of whole-exome sequencing for ambulant children with suspected monogenic conditions. JAMA Pediatr. 2017;171(9):855–862.
    View this article via: CrossRef PubMed Google Scholar
25. Boycott KM, Innes AM. When one diagnosis is not enough. N Engl J Med. 2017;376(1):83–85.
    View this article via: CrossRef PubMed Google Scholar
26. Posey JE, et al. Resolution of disease phenotypes resulting from multilocus genomic variation. N Engl J Med. 2017;376(1):21–31.
    View this article via: CrossRef PubMed Google Scholar
27. Pengelly RJ, et al. Evaluating phenotype-driven approaches for genetic diagnoses from exomes in a clinical setting. Sci Rep. 2017;7(1):13509.
    View this article via: CrossRef PubMed Google Scholar
28. Smedley D, Robinson PN. Phenotype-driven strategies for exome prioritization of human Mendelian disease genes. Genome Med. 2015;7(1):81.
    View this article via: CrossRef PubMed Google Scholar
29. Zemojtel T, et al. Effective diagnosis of genetic disease by computational phenotype analysis of the disease-associated genome. Sci Transl Med. 2014;6(252):252ra123.
    View this article via: PubMed Google Scholar
30. Javed A, et al. Phen-Gen: combining phenotype and genotype to analyze rare disorders. Nat Methods. 2014;11(9):935–937.
    View this article via: CrossRef PubMed Google Scholar
31. Burdick KJ, et al. Limitations of exome sequencing in detecting rare and undiagnosed diseases. Am J Med Genet A. 2020;182(6):1400–1406.
    View this article via: CrossRef PubMed Google Scholar
32. Johns Hopkins University. Online Mendelian Inheritance in Man. https://omim.org Updated January 24, 2022. Accessed January 25, 2022.
33. Hoekstra AS, et al. Simple and rapid characterization of novel large germline deletions in SDHB, SDHC and SDHD-related paraganglioma. Clin Genet. 2017;91(4):536–544.
    View this article via: CrossRef PubMed Google Scholar
34. Cope H, et al. Missed diagnoses: clinically relevant lessons learned through medical mysteries solved by the Undiagnosed Diseases Network. Mol Genet Genomic Med. 2020;8(10):e1397.
    View this article via: PubMed Google Scholar
35. Manning M, et al. Array-based technology and recommendations for utilization in medical genetics practice for detection of chromosomal abnormalities. Genet Med. 2010;12(11):742–745.
    View this article via: CrossRef PubMed Google Scholar
36. Manickam K, et al. Exome and genome sequencing for pediatric patients with congenital anomalies or intellectual disability: an evidence-based clinical guideline of the American College of Medical Genetics and Genomics (ACMG). Genet Med. 2021;23(11):2029–2037.
    View this article via: CrossRef PubMed Google Scholar
37. Cassini TA, et al. Whole genome sequencing reveals novel IGHMBP2 variant leading to unique cryptic splice-site and Charcot-Marie-Tooth phenotype with early onset symptoms. Mol Genet Genomic Med. 2019;7(6):e00676.
    View this article via: PubMed Google Scholar
38. Smith RB, et al. Clinical diagnosis of neurofibromatosis type I in multiple family members due to cosegregation of a unique balanced translocation with disruption of the NF1 locus: testing considerations for accurate diagnosis. Am J Med Genet A. 2021;185(4):1222–1227.
    View this article via: CrossRef PubMed Google Scholar
39. Boxer AL, et al. Clinical, neuroimaging and neuropathological features of a new chromosome 9p-linked FTD-ALS family. J Neurol Neurosurg Psychiatry. 2011;82(2):196–203.
    View this article via: CrossRef PubMed Google Scholar
40. DeJesus-Hernandez M, et al. Expanded GGGGCC hexanucleotide repeat in noncoding region of C9ORF72 causes chromosome 9p-linked FTD and ALS. Neuron. 2011;72(2):245–256.
    View this article via: CrossRef PubMed Google Scholar
41. Renton AE, et al. A hexanucleotide repeat expansion in C9ORF72 is the cause of chromosome 9p21-linked ALS-FTD. Neuron. 2011;72(2):257–268.
    View this article via: CrossRef PubMed Google Scholar
42. Bahlo M, et al. Recent advances in the detection of repeat expansions with short-read next-generation sequencing. F1000Res. 2018;7(F1000 Faculty Rev):736.
    View this article via: PubMed Google Scholar
43. Rafehi H, et al. Bioinformatics-based identification of expanded repeats: a non-reference intronic pentamer expansion in RFC1 causes CANVAS. Am J Hum Genet. 2019;105(1):151–165.
    View this article via: CrossRef PubMed Google Scholar
44. Rajan-Babu IS, et al. Genome-wide sequencing as a first-tier screening test for short tandem repeat expansions. Genome Med. 2021;13(1):126.
    View this article via: CrossRef PubMed Google Scholar
45. Shendure J, et al. DNA sequencing at 40: past, present and future. Nature. 2017;550(7676):345–353.
    View this article via: CrossRef PubMed Google Scholar
46. Logsdon GA, et al. Long-read human genome sequencing and its applications. Nat Rev Genet. 2020;21(10):597–614.
    View this article via: CrossRef PubMed Google Scholar
47. 1000 Genomes Project Consortium, et al. A global reference for human genetic variation. Nature. 2015;526(7571):68–74.
    View this article via: CrossRef PubMed Google Scholar
48. Chaisson MJP, et al. Multi-platform discovery of haplotype-resolved structural variation in human genomes. Nat Commun. 2019;10(1):1784.
    View this article via: CrossRef PubMed Google Scholar
49. Miller DE, et al. Targeted long-read sequencing identifies missing disease-causing variation. Am J Hum Genet. 2021;108(8):1436–1449.
    View this article via: CrossRef PubMed Google Scholar
50. Zhao X, et al. Expectations and blind spots for structural variation detection from long-read assemblies and short-read genome sequencing technologies. Am J Hum Genet. 2021;108(5):919–928.
    View this article via: CrossRef PubMed Google Scholar
51. Zhang JY, et al. Using de novo assembly to identify structural variation of eight complex immune system gene regions. PLoS Comput Biol. 2021;17(8):e1009254.
    View this article via: CrossRef PubMed Google Scholar
52. Wenger AM, et al. Accurate circular consensus long-read sequencing improves variant detection and assembly of a human genome. Nat Biotechnol. 2019;37(10):1155–1162.
    View this article via: CrossRef PubMed Google Scholar
53. Mitsuhashi S, Matsumoto N. Long-read sequencing for rare human genetic diseases. J Hum Genet. 2020;65(1):11–19.
    View this article via: CrossRef PubMed Google Scholar
54. Amarasinghe SL, et al. Opportunities and challenges in long-read sequencing data analysis. Genome Biol. 2020;21(1):30.
    View this article via: CrossRef PubMed Google Scholar
55. Rehm HL, et al. ClinGen—the Clinical Genome Resource. N Engl J Med. 2015;372(23):2235–2242.
    View this article via: CrossRef PubMed Google Scholar
56. Brokamp E, et al. One is the loneliest number: genotypic matchmaking using the electronic health record. Genet Med. 2021;23(10):1830–1832.
    View this article via: CrossRef PubMed Google Scholar
57. Sobreira NLM, et al. Matchmaker exchange. Curr Protoc Hum Genet. 2017;95:9.31.1–9.31.15.
    View this article via: PubMed Google Scholar
58. Azzariti DR, Hamosh A. Genomic data sharing for novel mendelian disease gene discovery: the matchmaker exchange. Annu Rev Genomics Hum Genet. 2020;21:305–326.
    View this article via: CrossRef PubMed Google Scholar
59. Sobreira N, et al. GeneMatcher: a matching tool for connecting investigators with an interest in the same gene. Hum Mutat. 2015;36(10):928–930.
    View this article via: CrossRef PubMed Google Scholar
60. Roden DM, et al. Development of a large-scale de-identified DNA biobank to enable personalized medicine. Clin Pharmacol Ther. 2008;84(3):362–369.
    View this article via: CrossRef PubMed Google Scholar
61. Langmead B, Nellore A. Cloud computing for genomic data analysis and collaboration. Nat Rev Genet. 2018;19(4):208–219.
    View this article via: CrossRef PubMed Google Scholar
62. All of Us Research Program Investigators, et al. The “All of Us” Research Program. N Engl J Med. 2019;381(7):668–676.
    View this article via: CrossRef PubMed Google Scholar
63. Freed AS, et al. The impact of rapid exome sequencing on medical management of critically ill children. J Pediatr. 2020;226:202–212.
    View this article via: CrossRef PubMed Google Scholar
64. Gubbels CS, et al. Prospective, phenotype-driven selection of critically ill neonates for rapid exome sequencing is associated with high diagnostic yield. Genet Med. 2020;22(4):736–744.
    View this article via: CrossRef PubMed Google Scholar
65. Meng L, et al. Use of exome sequencing for infants in intensive care units: ascertainment of severe single-gene disorders and effect on medical management. JAMA Pediatr. 2017;171(12):e173438.
    View this article via: CrossRef PubMed Google Scholar
66. Stark Z, et al. A prospective evaluation of whole-exome sequencing as a first-tier molecular test in infants with suspected monogenic disorders. Genet Med. 2016;18(11):1090–1096.
    View this article via: CrossRef PubMed Google Scholar
67. Dimmock DP, et al. An RCT of rapid genomic sequencing among seriously ill infants results in high clinical utility, changes in management, and low perceived harm. Am J Hum Genet. 2020;107(5):942–952.
    View this article via: CrossRef PubMed Google Scholar
68. French CE, et al. Whole genome sequencing reveals that genetic conditions are frequent in intensively ill children. Intensive Care Med. 2019;45(5):627–636.
    View this article via: CrossRef PubMed Google Scholar
69. Petrikin JE, et al. Rapid whole genome sequencing and precision neonatology. Semin Perinatol. 2015;39(8):623–631.
    View this article via: CrossRef PubMed Google Scholar
70. Willig LK, et al. Whole-genome sequencing for identification of Mendelian disorders in critically ill infants: a retrospective analysis of diagnostic and clinical findings. Lancet Respir Med. 2015;3(5):377–387.
    View this article via: CrossRef PubMed Google Scholar
71. Sanford EF, et al. Rapid whole genome sequencing has clinical utility in children in the PICU. Pediatr Crit Care Med. 2019;20(11):1007–1020.
    View this article via: CrossRef PubMed Google Scholar
72. Dimmock D, et al. Project Baby Bear: rapid precision care incorporating rWGS in 5 California children’s hospitals demonstrates improved clinical outcomes and reduced costs of care. Am J Hum Genet. 2021;108(7):1231–1238.
    View this article via: CrossRef PubMed Google Scholar
73. Petrikin JE, et al. The NSIGHT1-randomized controlled trial: rapid whole-genome sequencing for accelerated etiologic diagnosis in critically ill infants. NPJ Genom Med. 2018;3:6.
    View this article via: CrossRef PubMed Google Scholar
74. Splinter K, et al. Effect of genetic diagnosis on patients with previously undiagnosed disease. N Engl J Med. 2018;379(22):2131–2139.
    View this article via: CrossRef PubMed Google Scholar
75. Palmer EE, et al. Integrating exome sequencing into a diagnostic pathway for epileptic encephalopathy: evidence of clinical utility and cost effectiveness. Mol Genet Genomic Med. 2018;6(2):186–199.
    View this article via: CrossRef PubMed Google Scholar
76. Dragojlovic N, et al. The cost and diagnostic yield of exome sequencing for children with suspected genetic disorders: a benchmarking study. Genet Med. 2018;20(9):1013–1021.
    View this article via: PubMed Google Scholar
77. Beyter D, et al. Long-read sequencing of 3,622 Icelanders provides insight into the role of structural variants in human diseases and other traits. Nat Genet. 2021;53(6):779–786.
    View this article via: CrossRef PubMed Google Scholar
78. Sedlazeck FJ, et al. Accurate detection of complex structural variations using single-molecule sequencing. Nat Methods. 2018;15(6):461–468.
    View this article via: CrossRef PubMed Google Scholar
79. Chorin O, et al. Transcriptome sequencing identifies a noncoding, deep intronic variant in CLCN7 causing autosomal recessive osteopetrosis. Mol Genet Genomic Med. 2020;8(10):e1405.
    View this article via: PubMed Google Scholar
80. Hartley T, et al. New diagnostic approaches for undiagnosed rare genetic diseases. Annu Rev Genomics Hum Genet. 2020;21:351–372.
    View this article via: CrossRef PubMed Google Scholar
81. Grant P, et al. Out-of-pocket and private pay in clinical genetic testing: a scoping review. Clin Genet. 2021;100(5):504–521.
    View this article via: CrossRef PubMed Google Scholar
82. Zhong A, et al. Ethical, social, and cultural issues related to clinical genetic testing and counseling in low- and middle-income countries: protocol for a systematic review. Syst Rev. 2017;6(1):140.
    View this article via: CrossRef PubMed Google Scholar
83. Manolio TA, et al. Global implementation of genomic medicine: we are not alone. Sci Transl Med. 2015;7(290):290ps13.
    View this article via: PubMed Google Scholar
84. Paul JL, et al. A theory-informed systematic review of clinicians’ genetic testing practices. Eur J Hum Genet. 2018;26(10):1401–1416.
    View this article via: CrossRef PubMed Google Scholar