Collaborative projects, data sharing, and building disease cohorts have proven invaluable in genomics. In 2010, MLL2 (KMT2D) was discovered as the cause of the Kabuki syndrome (MIM: 147920). Ten unrelated patients with the same characteristic clinical phenotype underwent exome sequencing. Seven of the 10 individuals were found to have loss-of-function variants in MLL2, which led to its disease association (Ng et al., 2010). Historically, the approach of building a case series of affected individuals has been a rate-limiting step, relying on local connections or collaborations built through conferences or publications. Given the rarity of monogenic disorders, it can take many years to accrue sufficiently sized cohorts with similar clinical features and genotypes. This method is therefore inefficient and inadequate to rapidly support novel gene discovery (Azzariti and Hamosh, 2020).

In 2017, the directors’ board of the American College of Medical Genetics and Genomics released a position statement on how genomic data sharing is critical to improving genetic healthcare (ACMG Board of Directors, 2017). With an ever more connected world, global efforts to share genotype and phenotype data have proven essential in the endeavor of novel gene discovery. Improved data governance, drives for open data science, and advancing informatics methods have since led to the practice of genomic matchmaking, facilitating researchers and clinicians from across the globe to share phenotype/genotype data for accelerated discoveries (Azzariti and Hamosh, 2020).

In 2015, the Matchmaker Exchange (MME) was launched, providing a systematic and robust approach to novel Mendelian gene discovery by facilitating a mechanism for matching patients across genomic centers, research laboratories, diagnostic laboratories, and physicians through a federated network (Figure 1) (Philippakis et al., 2015).

MME builds on the success of earlier genomic matchmaking platforms by connecting datasets through an application programming interface (API) enabling searches of multiple databases with a single query. The advantage of using a federated network enables individual submitters to maintain control and autonomy over their data and keep the content up to date, while ensuring compliance with their local and national data-sharing policies (Azzariti and Hamosh, 2020). By identifying additional affected kindreds with overlapping phenotypes, the best candidate variants and genes can be targeted for functional validation. The MME API has been widely adopted by scientists and clinicians globally and has led to numerous international collaborations and publications. One such example is the discovery of a KMT2E-related neurodevelopmental disorder, the O’Donnell-Luria-Rodan syndrome (MIM: 618512), following identification of 38 individuals from 36 families, of which 28 were ascertained using MME. This discovery goes beyond elucidating disease–gene etiology and has identified a potential therapy already widely used in healthcare that could be evaluated for this syndrome (O’Donnell-Luria et al., 2019).

It could be argued that no one is more invested in ending the diagnostic odyssey of rare disease than affected individuals and their families. In an era when patients are actively involved in research studies as participants (Kaye et al., 2012), it is unsurprising that patients and caregivers are also invested in genomic matchmaking efforts (Lambertson et al., 2015). Patients and families are beginning to take control of their own data and utilize open data sharing and social media in an effort to discover new genetic disorders. Embedded within the MME API is a family-facing platform called MyGene2, which gives patients and caregivers autonomy over their data, facilitating direct data sharing when desired, while still enabling scientists and clinicians to access these shared anonymized data (Azzariti and Hamosh, 2020).

Social networking sites such as Facebook, Twitter, and Instagram are also proving popular with patients/caregivers as a matchmaking resource (Macnamara et al., 2019). In 2014, Matthew and Cristina Might harnessed the power of social media to identify additional cases of NGLY1 deficiency, leading to identification of a new gene disorder (Might and Wilsey, 2014; Might and Might, 2017). Following their son’s diagnosis, the Might family explored options for conceiving a child unaffected by the same condition. Their son’s diagnosis facilitated not only conception of a healthy sibling but a pathway for other affected families to conceive healthy children using preimplantation genetic testing or non-invasive prenatal diagnostic testing (Might and Might, 2017). The Might family created a legacy for others to follow, having built a global community of families providing mutual support, in addition to facilitating research and international NGLY1 meetings (Might and Wilsey, 2014).

Inspired by the success of the Might family, families across the globe have harnessed the networking potential of social media to match with other affected kindreds with similar phenotypes and genotypes. Indeed, social media additionally facilitated the identification of three children with variants of uncertain significance in KDM1A, leading to discovery of another novel gene disorder (MIM: 616728) (Chong et al., 2016). The success of such endeavors has now inspired the Undiagnosed Diseases Network (UDN), started at the National Institute for Health in 2008, with 11 additional clinical sites across the United States, to use social media in a similar way. With appropriate consent, webpages are created for individual participants, showcasing the clinical phenotype, significant variants, and candidate genes. This approach has proven successful in identifying additional affected patients with variants in NACC1, leading to the discovery of its associated phenotype (Macnamara et al., 2019).

Genomic sequencing has become increasingly affordable and possible in recent years. However, exome and genome sequencing are seldom first-line investigations, with many healthcare systems and health insurance policies not covering the cost. This has perhaps inspired the creation of large-scale international sequencing programs, often with government funding, offering exome and/or genome sequencing to thousands of rare disease patients and their families. These projects benefit from pooled resources and focus on diagnosing patients who were undiagnosed following conventional clinical testing, in addition to better elucidating the underlying mechanisms of Mendelian diseases. Such examples include the United Kingdom’s 100,000 Genomes Project (100KGP) (Turnbull et al., 2018) and Deciphering Developmental Disorders study (Firth et al., 2011) as well as the United States’ Centers for Mendelian Genomics (Bamshad et al., 2012). These programs benefit from sequencing large numbers of patients with improved power to match patients with similar genotypes and phenotypes, both internally and through the MME (Philippakis et al., 2015). Furthermore, most of these programs recruit patients for both clinical diagnostics and follow-on research, meaning that where possible, novel discoveries and variants of uncertain significance can be investigated further; this has previously been a limitation of clinical diagnostic studies (Seaby and Ennis, 2020). Consequently, thousands of novel gene discoveries have been identified through these projects. Indeed, by 2018, the Centers of Mendelian Genomics had reported >3,500 disease gene–phenotype pairs, expanding both known and novel disease gene associations (Posey et al., 2019). And in 2020, 28 new genetic disorders were discovered by leveraging data from the Deciphering Developmental Disorders study (Kaplanis et al., 2020).

In recent years, novel gene discovery has shifted from phenotype-driven methods to genotype-driven approaches, i.e., taking genotype data and matching phenotypes to that genotype through matchmaking efforts, though both remain important (Bamshad et al., 2019). Efforts to standardize phenotype terms through the Human Phenotype Ontology (HPO) database have aided comparative statistics using a universal library of agreed clinical terms involved in disease (Robinson et al., 2008). This has paved the way for computational phenotype analyses that can assess a candidate gene’s relevance to phenotype data observed in patient(s). A number of tools (Table 1) have been developed that estimate the similarity between HPO terms in an individual and those representing disease in a database. By incorporating phenotype ontology data across species, these tools are capable of prioritizing candidate genes without known disease association (Köhler et al., 2009; Schulz et al., 2011; Bauer et al., 2012). Similar approaches have been commercialized, taking advantage of advanced artificial intelligence to identify and rank potential disease-causing variants following a multidimensional analysis; examples include Fabric Genomics¹ and Emedgene². More experience and data are needed to understand the strengths and limitations of these tools.

One of the challenges for novel gene discovery is the requirement for accurate and deep phenotyping. Optimally, this should be collected longitudinally (Seaby and Ennis, 2020). While HPO terms do help to standardize the recording of phenotype information and indeed are used universally in many databases including those connected through MME (Philippakis et al., 2015), they are often only collected at a point in time and may lack the “full narrative” of the clinical history. This can be problematic when assessing new genotype–phenotype correlations, since for many neurodevelopmental disorders, phenotypes can significantly overlap. It can also be difficult to weight the severity or prominence of clinical features as the conversion to a list of terms tends to weight all the features similarly.

Given the mutation rate and the Earth’s current population size, we expect to observe every variant compatible with life in a living human. Indeed, the aggregation of large population datasets has begun to reveal the spectrum of damaging variants across the human genome (Lek et al., 2016; Karczewski et al., 2020). It is typical to observe approximately 100 loss-of-function variants per genome with ∼20 genes completely inactivated (knockouts) even in perfectly healthy individuals from the general population (MacArthur et al., 2012). Population data can be utilized to evaluate the strength of natural selection at the gene level and to differentiate rare from common loss-of-function variants. As deleterious variants are purged from human populations through natural selection, there are opportunities to identify genes and regions that are constrained for variation compared to expected mutation rates, revealing which genes are most intolerant to inactivation of one (haploinsufficient) or both (knockout) copies (Samocha et al., 2014; Karczewski et al., 2020).

Lek et al. (2016) (Kosmicki et al., 2017) defined a set of genes with high probability of intolerance to heterozygous predicted loss-of-function variation (pLI) modeled on ∼60,000 exomes from the Exome Aggregation Consortium (ExAC) population database (Lek et al., 2016). This pLI score can be used to identify candidate haploinsufficient disease genes constrained for loss of function in a dichotomous way; i.e., a gene is predicted to be haploinsufficient (pLI > 0.9) or not (Figure 2).

Karczewski et al. (2020) refined the model and regenerated pLI scores utilizing a larger dataset of ∼141,000 exomes and genomes from the Genome Aggregation Database (gnomAD)³. The authors also developed the Lower Observed/Expected Upper-bound Fraction (LOEUF) score, a continuous metric which places >19,000 human genes on a spectrum of intolerance to inactivation (Figure 2). Genes with the lowest LOEUF scores, i.e., the fewest loss-of-function variants compared to an expectation, are the most constrained for loss of function, highlighting their potential biological essentiality. Both LOEUF and pLI were validated by comparison to several orthogonal indicators of constraint and shown to be accurate at discriminating haploinsufficient disease genes from autosomal recessive and polymorphic (unconstrained) genes (Karczewski et al., 2020). A companion paper by Collins et al. (2020) additionally showed that structural variants share the same pattern of constraint as LOEUF and are responsible for about a quarter of all rare loss-of-function events per genome.

As LOEUF identifies genes constrained for loss-of-function variation, we expect these genes to be enriched for dominant disease genes and to lesser extent recessive disease genes. As of January 2021, 65% of genes in the lowest LOEUF decile are yet to have an OMIM disease association (calculated using data from https://omim.org), highlighting thousands of high-probability candidate disease genes awaiting discovery of the associated phenotypes.

The majority of coding variants of uncertain clinical significance are missense variants, as reported in ClinVar, a public database where diagnostic laboratories and researchers share variant classifications (i.e., pathogenic, benign, and uncertain significance) (Pérez-Palma et al., 2019). Similar to methods for assessing loss-of-function constraint, methods to identify missense constraint have emerged by comparing the observed vs. expected numbers of missense variants modeled on population data (Lek et al., 2016; Samocha et al., 2017; Havrilla et al., 2019; Hayeck et al., 2019; Pérez-Palma et al., 2020). However, missense constraint varies across a gene; for example, unstructured regions are often less constrained than important functional domains, which has necessitated the development of regional missense constraint models (Havrilla et al., 2019; Abramovs et al., 2020). Furthermore, clustering patterns of pathogenic missense variants vary depending on the inheritance pattern. Turner et al. (2015) showed that dominant missense variants cluster more than recessive variants. Therefore, testing for non-random clustering patterns may identify novel regions of interest across large sample sizes (Turner et al., 2015). The application of these metrics has aided in the discovery of new gene disorders, including a de novo missense variant in a constrained region of GABRA2 responsible for an early-onset epileptic encephalopathy (MIM: 618557) (Orenstein et al., 2018).

Identifying the phenotypic effects of gene disruption may be possible using model organisms when there is enough conserved evolutionary function of the pathway/organ/system involving the gene of interest. Where the functional consequences of most human gene variants are yet to be established, model organism databases serve as a useful resource. Indeed, 58% of human genes have orthologs with disease-associated phenotypes reported in at least one model organism (Mungall et al., 2017). Although non-human models are not necessarily perfect proxies for human diseases, they can still serve as important biological models, particularly when data are aggregated across species.

The Monarch Initiative⁴ is an open-science, collaborative project that aims to integrate phenotype–genotype data from a variety of species and sources (Mungall et al., 2017). Its user-friendly web portal promotes rapid assessment of phenotypes of orthologs in organisms and other species. Researchers can query genes, phenotypes, and diseases to identify candidate disease genes. Exomiser (Smedley et al., 2015) and Genomiser (Smedley et al., 2016) have utilized the Monarch Initiative in their gene prediction algorithms, which have led to diagnoses in participants in the UDN including the aforementioned discovery that the disruption of STIM1 results in the York platelet syndrome (MIM: 805070) (Bone et al., 2016).

Several mouse model organism databases exist including the Mouse Genome Database (MGD) (Smith et al., 2017; Bult et al., 2019), the Knockout Mouse Project (KOMP) (Austin et al., 2004), and the International Mouse Phenotyping Consortium (IMPC) (Meehan et al., 2017; Muñoz-Fuentes et al., 2018). These projects are building comprehensive catalogs of mammalian gene function, genotype–phenotype associations, and detailed phenotype data from mouse knockouts of every protein-coding gene (Muñoz-Fuentes et al., 2018; Bult et al., 2019). By 2019, the IMPC has fully or partially phenotyped 5,861 mouse genes, a third of which are non-viable (Muñoz-Fuentes et al., 2018; Cacheiro et al., 2019). Data from IMPC have aided the discovery of many novel Mendelian phenotypes (Bowl et al., 2017; Moore et al., 2018; Rozman et al., 2018). That said, there is still much more to be gleaned from mouse data; of the >10,000 mouse genes linked to at least one non-lethal phenotype in a mutant strain in MGD, Bamshad et al. (2019) showed that human orthologs for 72% of those genes are yet to be associated with a Mendelian disorder, providing another rich data source for candidate genes awaiting discovery of the human Mendelian phenotype.

Identifying novel disease genes can be challenged by incomplete penetrance, that is to say, when a disease-causing variant does not always result in any clinical expression of the disease. If a novel candidate gene has been associated with a given phenotype, yet some or all alleles are incompletely penetrant, then it can be difficult to gather sufficient evidence for a new disease–gene association using traditional genetic evidence such as case observations and familial segregation. To mitigate this, larger cohorts that can support statistical association studies must be pursued. Furthermore, researchers are exploring how combinations of genomic variants such as oligogenic models or co-inherited protective alleles, environmental exposures, and mosaicism may impact the onset of Mendelian disorders (Gruber and Bogunovic, 2020). One such approach is to specifically identify individuals that are resilient to rare disease, despite harboring pathogenic variants (Chen et al., 2016). Another area of interest is how cis-regulatory variation may modify the penetrance of coding variants (Castel et al., 2018).

In recent years, government funding has invested in national sequencing projects for rare disease (Turnbull et al., 2018; Posey et al., 2019). Despite 100,000 individuals with rare disease being sequenced in the United Kingdom as part of the Genomics England 100KGP, the diagnostic rates are similar to those reported elsewhere in the literature (Houses of Parliament, 2015; Turnbull et al., 2018). However, the scale of such datasets welcomes opportunities for new approaches to novel gene discovery.

With increasing data available on genetic variation from a variety of sources including gene constraint, mouse models, and phenotype-driven methods, there is scope to utilize the power of large cohort sizes for novel gene discovery. Instead of bringing a patient to a gene, there are opportunities, with large enough sample sizes, to be sufficiently powered to detect rare variation and bring candidate genes to large genomic datasets from patients (Figure 3). These “gene-to-patient” approaches are already being applied to accelerate novel gene discovery and prioritize genes for functional studies (Seaby and Ennis, 2020).

Multiple omics technologies such as epigenomics, transcriptomics, metabolomics, microbiomics, and proteomics are being adopted as approaches in the effort to delineate the functional impact of genetic variation (Figure 4) (Hasin et al., 2017). These integrative approaches can complement genomic data and aid in the validation and discovery of novel genes.

The Genotype-Tissue Expression (GTEx) project⁵ (Lonsdale et al., 2013) provides a public repository of tissue-specific gene expression and a multi-tissue reference for identifying variants associated with changes in gene expression or expression quantitative trait loci (eQTL) (Stranger et al., 2017). The GTEx consortium recently published results on their v8 release, providing insights into functional mechanisms and the architecture of genetic regulation (GTEx Consortium, 2020). The integration of transcriptome sequencing (RNA-seq) has led to improved diagnostic rates for Mendelian diseases (Kremer et al., 2017; Wai et al., 2020), even when no strong candidate variants were identified from exome or genome data (Cummings et al., 2017). Expression outliers, altered splicing, and allelic imbalance in the transcriptome due to non-sense-mediated decay can all be clues to candidate genes worth closer scrutiny in the exome or genome data (Brechtmann et al., 2018; Green et al., 2018; Mertes et al., 2021). Large-scale transcriptome data can also be used in network analysis (Deelen et al., 2019). One caveat with RNA-seq is that splicing aberrations and differential gene expression are best assessed by sampling disease-relevant tissues. However, these may not always be clinically accessible, e.g., brain tissue in neurodevelopmental disorders. Aicher et al. (2020) showed that many splicing events in non-clinically accessible tissues are lowly expressed and poorly evaluated from more commonly accessible tissues such as skin and blood. The authors developed a tool (MAJIQ-CAT) which allows researchers to explore potentially accessible tissues that best represent splicing in genes of interest (Aicher et al., 2020). A recent preprint in 2021 describes an alternative approach that has advantages over expression-based methods. Minimum Required Sequencing Depth (MRSD) informs biosample selection (whole blood, lymphoblastic cell lines, and skeletal muscle) by estimating the minimum sequencing depth required from RNA-sequencing to achieve desired coverage across a given gene or gene panel. The authors reported high precision, and their results suggest that lymphoblastic cell lines may be suitable for ∼70% of established disease gene panels (Rowlands et al., 2021).

The Encyclopedia of DNA Elements (ENCODE) and Roadmap Epigenomics projects have been instrumental in the generation of human reference epigenomes and epigenome maps, mainly from cell lines (Bernstein et al., 2010; Dunham et al., 2012; Satterlee et al., 2019). These data have successfully been used to conduct research on how the epigenome contributes to human development, environmental factors, and disease mechanisms (Dunham et al., 2012; Satterlee et al., 2019). More specifically, one of most commonly studied epigenetic phenomena, DNA methylation, is aiding diagnosis and gene discovery. Alterations in DNA methylation patterns are implicated in imprinting disorders and diseases of short tandem repeat (STR) expansions. The application of DNA methylation analyses has been successful in identifying molecular diagnoses in neurodevelopmental disorders where clinical microarray and other conventional genetic testing have been non-diagnostic (Aref-Eshghi et al., 2019; Turinsky et al., 2020). LaCroix et al. (2019) investigated cases of Baratela–Scott syndrome (BSS) (MIM: 615777) and identified hypermethylation of exon 1 of XYLT1 associated with a GGC expansion and gene silencing. This not only confirmed BSS as a trinucleotide repeat expansion disorder but also highlighted the relative prevalence of methylation abnormalities in the disease pathogenesis of BSS. The hypermethylated allele accounted for 50% of the pathogenic alleles in their cohort, showcasing the importance of investigating epigenetic changes in disease cohorts with missing heritability (LaCroix et al., 2019).

National biobanks such as the United Kingdom Biobank (Bycroft et al., 2018), United States All of Us Research Program (Sankar and Parker, 2017), and Finland Biobank (FinnGen) provide opportunities to study genomic data and phenotype data alongside associated molecular markers from electronic medical records. While their data are best studied in the context of complex disease, they are also important in rare disease by providing population-level allele frequencies, biomarker results, and phenotypic information for comparative analyses. Unlu et al. (2020) utilized Vanderbilt’s biobank BioVU to identify a phenotypic profile that aided in the identification of a novel Mendelian syndrome CATIFA (cleft lip, cataract, tooth abnormality, intellectual disability, facial dysmorphism, attention-deficit hyperactivity disorder) that is due to loss of function of RIC1 (MIM: 618761).

The emerging application of metabolomics with exome/genome sequencing is helping to improve diagnostic rates in rare disease. Targeted and untargeted metabolomics are proving successful in validating variants of uncertain significance in inborn errors of metabolism (Graham et al., 2018; Almontashiri et al., 2020). It is hoped that with increasing research, metabolomics will continue to complement rich genomic data and aid in discovery of novel genes.

These aforementioned approaches often applied in combination have been pivotal both in clinical diagnostics and in identification of novel candidate disease genes. For example, a study in 2015 using epigenomics, comparative genomics, and genome editing identified a pathway for adipocyte thermogenesis regulation involving IRX3, IRX5, and ARID5B in obesity (Claussnitzer et al., 2015), and in 2017, the complex I assembly factor TIMMDC1 was established as a novel mitochondrial disease–gene by utilizing genomic and transcriptomic sequencing (Kremer et al., 2017).

It is unquestionable that open science and data sharing have been pivotal in uplifting diagnoses and advancing the field of genomic medicine. International collaborations are now commonplace in matching patients across the globe with specific genotypes, leading to high-impact publications on novel gene discoveries (Philippakis et al., 2015; Azzariti and Hamosh, 2020). Furthermore, the use of variant databases such as ClinVar⁶ have proven invaluable in providing the scientific community with a repository of variants, classified by pathogenicity, that can be applied to variant analysis for diagnostic interpretation (Landrum et al., 2014).

The success of open data science has been further driven by cloud computing. The presence of large datasets on cloud platforms can facilitate the access of desired data within a secure data-sharing platform. Examples include NHGRI’s Genomic Data Science Analysis, Visualization, and Informatics Lab (AnVIL) Space⁷ where rare disease data from the Centers for Mendelian Genomics along with data from additional projects such as 1,000 Genomes, Centers for Common Disease Genomics, and GTEx can be accessed after application in dbGaP. Other trusted research environments include NHLBI’s BioData Catalyst cloud platform with TOPMed data; Genomics England Research environment where 100 KGP data are stored and accessed; and RD-Connect with rare disease genomic data from various European sources. In these trusted research environments, increasingly large amounts of data can be aggregated; researchers can bring tools directly to the data and share these analysis workflows, saving time, expense, and security risks of moving and maintaining local copies of large genomic datasets. However, data sharing presents its challenges. There is still urgent need for an international code of conduct that provides clear, unified data-sharing rules across jurisdictions that comply with regional laws such as the European General Data Protection Regulation (GDPR) and the United States’ Health insurance Portability and Accountability Act (Phillips et al., 2020).

Our understanding of the genetic basis of rare disease is constantly changing with new genes and variation being linked to disease at a rapid pace. Given the direct application of these discoveries to the clinical diagnosis of rare disease in patients, guidance is needed for understanding what information is ready to be incorporated into clinical care and what mechanisms are needed to quickly translate that information into medical practice. The Clinical Genome Resource (ClinGen) has developed a systematic framework for evaluating genetic and functional evidence for disease–gene relationships, enabling their classification as definitive, strong, moderate limited, no human evidence, disputed, or refuted with respect to their reported role in disease (Strande et al., 2017). ClinGen supports Gene Curation Expert Panels that bring together international groups of disease and curation experts to evaluate gene–disease claims in their respective fields⁸. ClinGen’s efforts are combined with other public and private gene curation efforts and are accessible within the Gene Curation Coalition database⁹. Currently, it is recommended that a gene–disease relationship reach moderate classification before it is included on predefined diagnostic gene panels for specific conditions (Bean et al., 2020). However, when performing exome and genome approaches on individuals with rare disease, variation can be detected in genes that have not yet been linked to disease but may be strong candidates. Although practices vary between laboratories and countries, some professional standards recommend reporting these findings back to patients when there is a reasonable chance that new evidence may evolve over time to strengthen the gene–disease relationship, similar to the return of variants of uncertain significance in genes already linked to the patient’s condition (Richards et al., 2015; Rehder et al., 2021). This approach also allows patients to be partners in solving the causes of rare disease (Mnookin, 2014). It is hoped that such a framework will achieve global recognition and be universally adopted to ensure consistency in translating research findings into the clinic.

While novel gene discoveries widen the known functional repertoire of disease genes, the focus and drive are ultimately uplifting diagnosis rates and improving patient outcomes. There have been thousands of pivotal Mendelian discoveries throughout history, and each one is no more important than another, at least not for the families involved.

Since the discovery of the CFTR gene in 1989 (Kerem et al., 1989), we are now able to diagnose cystic fibrosis (MIM: 602421) rapidly, predict pancreatic functional status, and plan preventative care with modulator therapy (Ramsey et al., 2011; Farrell et al., 2020). In 2004, the discovery that hypermorphic or gain-of-function variants of PCSK9 cause familial hypercholesterolemia type III (MIM: 603776) (Timms et al., 2004) has led to the successful development and FDA approval of monoclonal antibodies against PCSK9, which are also used to treat non-familial forms of hypercholesterolemia (Blom et al., 2014; Roth et al., 2014; Cannon et al., 2015; Kereiakes et al., 2015). As more collaborative, cohort-based studies have emerged in the NGS era, many candidate genes have been discovered that have directly impacted treatment and clinical outcomes. In one study on neurometabolic disorders, whole-exome sequencing diagnosed 68% of patients and identified 11 novel candidate genes, leading to a targeted intervention in 44% of patients (Tarailo-Graovac et al., 2016).

Diagnosing Mendelian disorders as a direct result of novel gene discovery not only impacts the primary patient involved but their families and caregivers. Families of children with rare genetic diseases are adversely impacted by lack of peer support groups and psychological support as well as delays in diagnosis (Anderson et al., 2013). Parents of children with rare disorders have called for better education, reduction in avoidable diagnostic delays, and early access to interventions and treatments (Zurynski et al., 2017). For many, a genetic diagnosis can be life changing, even in the absence of a therapeutic option (Lingen et al., 2016). Following diagnosis, quality of life is often improved by participation in support groups that can provide longitudinal prognostic information, genetic counseling, and informed reproductive decisions with opportunities for pre-implantation genetic diagnosis or prenatal testing particularly in the case of inherited variants where there is a sizable recurrence risk (Sexton et al., 2008; Wojcik et al., 2020).

Undiagnosed rare diseases are hugely expensive. A typical patient’s diagnostic odyssey lasts an average of 8 years and costs a total of $5,000,000 throughout a patient’s lifetime (Chong et al., 2015). Two prospective Australian studies have shown that early exome sequencing is making significant headway as a cost-saving diagnostic approach. Stark et al. (2017) showed that integrating whole-exome sequencing as a first-line test had an incremental cost saving per additional diagnosis of (converted to United States dollars) $1,543 (95% CI: $92–4,143). The cost per diagnosis was $4,248 (95% CI: $3,425–5,588), $14,893 less than standard diagnostic care (Stark et al., 2017). Tan et al. (2017) concluded that whole-exome sequencing performed at initial presentation to tertiary care resulted in an incremental cost saving of (converted to United States dollars) $6,383 per additional diagnosis (95 CI: $3,045–10,900) compared with standard diagnostic care. However, cost savings are only possible when sequencing can identify the causal variant. Therefore, every new genetic disorder identified, published, and shared in publicly available databases will have wide-reaching diagnostic and cost-saving potential. Taking the average of costs saved per additional diagnosis from the two studies ($3,964) and extrapolating this on 100,000 patients could save an estimated US$400 million.