The Genetic Basis of Mendelian Phenotypes: Discoveries, Challenges, and Opportunities

Abstract

Discovering the genetic basis of a Mendelian phenotype establishes a causal link between genotype and phenotype, making possible carrier and population screening and direct diagnosis. Such discoveries also contribute to our knowledge of gene function, gene regulation, development, and biological mechanisms that can be used for developing new therapeutics. As of February 2015, 2,937 genes underlying 4,163 Mendelian phenotypes have been discovered, but the genes underlying ∼50% (i.e., 3,152) of all known Mendelian phenotypes are still unknown, and many more Mendelian conditions have yet to be recognized. This is a formidable gap in biomedical knowledge. Accordingly, in December 2011, the NIH established the Centers for Mendelian Genomics (CMGs) to provide the collaborative framework and infrastructure necessary for undertaking large-scale whole-exome sequencing and discovery of the genetic variants responsible for Mendelian phenotypes. In partnership with 529 investigators from 261 institutions in 36 countries, the CMGs assessed 18,863 samples from 8,838 families representing 579 known and 470 novel Mendelian phenotypes as of January 2015. This collaborative effort has identified 956 genes, including 375 not previously associated with human health, that underlie a Mendelian phenotype. These results provide insight into study design and analytical strategies, identify novel mechanisms of disease, and reveal the extensive clinical variability of Mendelian phenotypes. Discovering the gene underlying every Mendelian phenotype will require tackling challenges such as worldwide ascertainment and phenotypic characterization of families affected by Mendelian conditions, improvement in sequencing and analytical techniques, and pervasive sharing of phenotypic and genomic data among researchers, clinicians, and families.

Main Text

Introduction

Improved understanding of human disease was a primary goal of the Human Genome Project (HGP).¹ This promise has, in part, been realized with the identification of the consequence of germline mutation (single-nucleotide variants [SNVs] and copy-number variants [CNVs]) for more than 2,900 protein-coding genes in humans.^2–4 These disease-associated mutations directly link DNA variants to altered protein function or dosage and to human phenotypes, thus transforming our understanding of the basic biology of development and physiological homeostasis in health and disease. Indeed, much of what is known about the relationship between gene function and human phenotypes is based on the study of rare variants underlying Mendelian phenotypes. Furthermore, these discoveries have identified new preventative, diagnostic, and therapeutic strategies for a growing number of rare and common diseases.^5–8

Much remains to be learned. The HGP and subsequent annotation efforts have established that there are ∼19,000 predicted protein-coding genes in humans.^9,10 Nearly all are conserved across the vertebrate lineage and are highly conserved since the origin of mammals ∼150–200 million years ago,^11–13 suggesting that certain mutations in every non-redundant gene will have phenotypic consequences, either constitutively or in response to specific environmental challenges. The continuing pace of discovery of new Mendelian phenotypes and the variants and genes underlying them supports this contention.

Whereas protein-coding regions compose only about 1% of the human genome, the overwhelming majority of Mendelian phenotypes identified thus far result from altered function, localization, or presence of the encoded proteins. Furthermore, few Mendelian phenotypes appear to be caused exclusively by mutations outside coding regions.¹⁴ This is not only a matter of ascertainment bias, given that loci that are well mapped by unbiased analysis of linkage data prior to the discovery of underlying causes have yielded extremely high ratios in favor of variants that alter protein function.¹⁵ However, progress in the elucidation of promoters and tissue-specific regulatory elements by projects such as ENCODE¹⁶ and in linking perturbations in these elements to alterations of gene expression holds promise for the identification of new Mendelian phenotypes caused by non-coding mutations.¹⁷ Mendelian phenotypes for which mutations have not been discovered in coding regions or canonical splice sites are ideal candidates for such studies.

Our increasingly sophisticated ability to phenotype humans, coupled with differences in physiology between humans and model organisms, argues that primary mutation discovery in humans will remain crucial to progress.¹⁸ Moreover, now that a finite set of protein-coding genes has been defined, determining the phenotypic consequences of their variation represents a vital and attainable goal partly because of the advances in the production and analysis of whole-exome sequencing (WES) and whole-genome sequencing (WGS) data.^19–24 Each successful discovery will define potential diagnostic, preventive, and therapeutic opportunities for the corresponding diseases and illuminate normal biology and disease mechanisms.^25,26

The Burden of Mendelian Disease

In aggregate, clinically recognized Mendelian phenotypes compose a substantial fraction (∼0.4% of live births) of known human diseases, and if all congenital anomalies are included, ∼8% of live births have a genetic disorder recognizable by early adulthood.²⁷ This translates to approximately eight million children born worldwide each year with a “serious” genetic condition, defined as a condition that is life threatening or has the potential to result in disability.²⁸ In the US alone, Mendelian disorders collectively affect more than 25 million people and are associated with high morbidity, mortality, and economic burden in both pediatric and adult populations.^28,29 Birth defects, of which Mendelian phenotypes compose an unknown but most likely substantial proportion, are the most common cause of death in the first year of life, and each year, more than three million children under the age of 5 years die from a birth defect, and a similar number survive with substantial morbidity. Beyond the emotional burden, each child with a genetic disorder has been estimated to cost the healthcare system a total of $5,000,000 during their lifetime.^28,29

It remains a challenge to diagnose many Mendelian phenotypes by phenotypic features and conventional diagnostic testing. In a general clinical genetics setting, the diagnostic rate is ∼50%.³⁰ Across a broader range of rare diseases, diagnostic rates are even lower. For example, in the NIH Undiagnosed Disease Program, the diagnostic rate was, despite state-of-the-art evaluations, 34% in adults and 11% in children.³¹ Moreover, the time to diagnosis is often prolonged (the “diagnostic odyssey”); in a European survey of the time to diagnosis of eight rare diseases, including cystic fibrosis (MIM: 602421) and fragile X syndrome (MIM: 309550), 25% of families waited between 5 and 30 years for a diagnosis, and the initial diagnosis was incorrect in 40% of these families.³²

Molecular diagnostic rates for many Mendelian phenotypes are similarly low. For example, of 292 well-characterized categories of Mendelian phenotypes for which clinical testing is available (GeneReviews, July 2014; Table S1), a causal variant can be identified in only 52% of cases overall (Figure 1A). Diagnostic rates vary widely by phenotype and are inversely correlated with the level of genetic heterogeneity (Spearman correlation ρ = −0.155, p value = 0.008019; Figure 1B; Table S2). These observations are consistent with diagnostic rates of 25%–75% for many major categories of inherited conditions (e.g., kidney disease,³³ cardiomyopathy,³⁴ epilepsy,³⁵ etc.) The increasing availability of clinical WES testing holds promise for improving diagnostic yield, and although the diagnostic rate reported with this technology is currently only ∼25%–30%, it represents a substantially higher rate than the diagnostic yield of other genomic assays, such as karyotyping (<5%) and array comparative genomic hybridization (∼15%–20%).^36,37 Importantly, studies of the diagnostic efficacy of clinical WES show that a substantial fraction (25%–30%) of the diagnostic successes depends on recent progress in the discovery of genes underlying disease.^36,38 This observation highlights the value of continued research into the genetic basis of Mendelian phenotypes. Moreover, it predicts that diagnostic rates will continue to increase as we work toward a more complete catalog of the genes and variants responsible for human disease.

Figure 1.

Clinical Diagnostic Rates of Mendelian Conditions for which Gene(s) Have Been Identified

All Mendelian conditions, or phenotypic series, included are listed in GeneReviews and might be genetically heterogeneous (i.e., caused by mutations in one or more genes).

(A) Histogram of the percentage of individuals who had a Mendelian condition (x axis) and who received a corresponding molecular diagnosis from clinical testing. Collectively, for 292 Mendelian conditions, a causal variant could be identified in only ∼52% of affected subjects overall.

(B) Boxplots show the molecular diagnostic rate (y axis) for Mendelian conditions organized by the number of causal genes (x axis). The diagnostic rate per condition is inversely correlated with the level of genetic heterogeneity (Spearman correlation ρ = −0.155, p value = 0.008019).

The Scope of Mendelian-Disease-Related Phenotypes

Our knowledge of the diversity of Mendelian phenotypes is increasingly sophisticated, but substantial gaps remain.³⁹ Specifically, it is challenging to establish the number of Mendelian phenotypes that exist, to delineate new Mendelian phenotypes, to distinguish novel from known Mendelian phenotypes, to define what constitutes expansion of a known phenotype, and to develop metrics for assessing the relationships and diversity of phenotypes caused by variants in the same gene. These gaps are due, in part, to the diversity of “normal” human morphology and physiology, the challenge of defining normal versus abnormal, and the difficulty of setting limits for traits with quantitative distributions.^40,41 Moreover, biological variation is, in large part, not binary but rather described by partially overlapping distributions.^42–45 With these caveats in mind, we analyzed OMIM entries (February 2015) to identify 7,440 rare Mendelian phenotypes (7,315 monogenic; 125 chromosomal duplications and/or deletions). This number is not static; ∼300 new Mendelian phenotypes are added to OMIM each year, and this probably underestimates the number of phenotypes newly recognized each year. Delineation of new Mendelian phenotypes in populations worldwide is limited by a lack of infrastructure, resources, and expertise^46,47 Moreover, studies of model organisms show that the number and types of recognized phenotypes increase with expanding environmental challenges.⁴⁸ Therefore, to completely enumerate “all” human Mendelian phenotypes, it will be necessary to consider a more comprehensive span of environmental conditions and develop more-sophisticated tools to evaluate phenotype.⁴⁹

Studies in mice with engineered loss-of-function (LOF) mutations suggest that the majority of the gene knockouts compatible with survival to birth are associated with a recognizable phenotype, whereas ∼30% of gene knockouts lead to in utero or perinatal lethality.⁵⁰ Of the latter, it remains to be determined whether partial LOF mutations (i.e., hypomorphic alleles) or other classes of mutations (e.g., gain of function, dosage differences due to gene amplification,⁵¹ etc.) in the same genes might result in viable phenotypes. Nevertheless, given the high degree of evolutionary conservation between humans and mice, mutations in the majority of non-redundant human protein-coding genes are likely to result in Mendelian phenotypes, most of which remain to be characterized (Figure 2).

Figure 2.

Relationship between Human Protein-Coding Genes and Mendelian Phenotypes

Of approximately ∼19,000 protein-coding genes predicted to exist in the human genome, variants that cause Mendelian phenotypes have been identified in ∼2,937 (∼15.5%; orange squares). Genes underlying ∼643 Mendelian phenotypes (∼3.38%; gray squares) have been mapped but not yet identified. On the basis of analysis of knockout mouse models, LOF variants in up to ∼30% of genes (∼5,960; red squares) could result in embryonic lethality in humans. Note that the consequences of missense variants in these genes could be different. For a minimum of ∼52% of genes (∼10,330; blue squares), the impact in humans has not yet been determined. Collectively, ∼16,063 genes remain candidates for Mendelian phenotypes.

Relationship between Mendelian Disease Traits and Common Disease

The etiologies of common diseases, such as hypertension, coronary artery disease, diabetes, obesity, scoliosis, and autism, are heterogeneous and typically include a small subset of individuals with a monogenic condition underlying their diagnosis with a common disease.^52–55 The variants responsible for this small fraction of affected individuals rarely explain much of the genetic contribution to these common diseases,^56,57 but they are nevertheless often highly relevant to our understanding of more-general mechanisms of these conditions.⁵³ A classic example in cardiovascular disease research is the identification of the genetic basis of rare, monogenic forms of hypercholesterolemia, which provided critical insights into the relevance of lipid transport.⁵⁸ In turn, these findings have led to the development of new therapies for common, complex cardiovascular diseases by targeting the implicated genes and pathways.⁵³ Collectively, nearly 20% of genes implicated in Mendelian phenotypes also either contain or are nearest to a variant responsible for a genome-wide association study (GWAS) signal that achieves genome-wide significance for a complex trait (Figure 3A; Supplemental Material and Methods; Figure S1). In contrast, ∼15% of all genes overall underlie a Mendelian phenotype, suggesting that genes implicated in Mendelian phenotypes are enriched in GWAS signals. The fraction of genes that are found near GWAS signals and in which variants are responsible for Mendelian phenotypes is also positively correlated with the strength of association (Figure 3B). Widespread co-morbidity among Mendelian phenotypes and complex diseases provides further evidence that variation in genes that underlie Mendelian phenotypes plays a role in complex disease.⁵⁹

Figure 3.

Relationship between GWAS Signals and Genes Underlying Mendelian Phenotypes

(A) Plot of the fraction of GWAS-signal genes that are also implicated in Mendelian phenotypes (MPs). Each orange dot represents the proportion of GWAS signals that, in a sliding window of 500 GWAS signals, are mapped to a gene also known to underlie a Mendelian condition. In GWAS signals, approximately 26.6% of genes with the top 500 lowest p values underlie a Mendelian phenotype. In contrast, only 14.2% of genes overall are known to underlie a Mendelian phenotype, suggesting that GWAS signals are more likely to be enriched with genes implicated in Mendelian phenotypes. Varied colored dots represent the percentage of genes underlying a Mendelian phenotype in GWAS signals underlying different phenotypic categories as follows (of increasing percentages from bottom to top): 10% for reproductive traits (blue); 11% for respiratory traits (gold); 13% for autoimmune inflammatory traits (dark green); 16% for immunologic traits (blue); 17% for mental-health traits (teal); 19% for infectious-disease traits (gray); 21% for anthropometric traits (brown); 23% for cancer (red); 25% for cardiovascular traits (tan); 26% for metabolomics traits (yellow); 28% for pharmacogenetic traits (green); and 33% for musculoskeletal traits (blue).

(B) Cumulative plot of the proportion of GWAS signals in which a gene underlying a Mendelian phenotype (MP) was found (orange dots) and GWAS signals in which a gene underlying a Mendelian phenotype was not found (gray dots). At virtually every p value, a higher proportion of GWAS signals overlapped genes underlying Mendelian phenotypes.

Development of new therapeutics to address common diseases that constitute major public-health problems is limited by the ignorance regarding the fundamental biology underlying disease pathogenesis.⁶⁰ As a consequence, 90% of drugs entering human clinical trials fail, commonly because of a lack of efficacy and/or unanticipated mechanism-based adverse effects.⁶¹ Studies of families affected by rare Mendelian phenotypes segregating with large-effect mutations that increase or decrease risk for common disease can directly establish the causal relationship between genes and pathways and common diseases and identify targets likely to have large beneficial effects and fewer mechanism-based adverse effects when manipulated. For example, certain Mendelian forms of high and low blood pressure are due to mutations that cause increases and decreases, respectively, in renal salt reabsorption and net salt balance; these discoveries identified promising new therapeutic targets, such as KCNJ1 (potassium channel, inwardly rectifying, subfamily J, member 1 [MIM: 600359]), for which drugs are now in clinical trials. Understanding the role of salt balance in blood pressure has provided the scientific basis for public-health efforts in more than 30 countries to reduce heart attacks, strokes, and mortality by modest reduction in dietary salt intake.⁶² Similarly, understanding the physiological effects of CFTR (cystic fibrosis transmembrane conductance regulator [MIM: 602421]) mutations responsible for cystic fibrosis has led to allele-specific therapies that significantly improve pulmonary function in affected individuals.⁶³ Other common-disease drugs based on gene discoveries for Mendelian phenotypes (e.g., orexin antagonists for sleep,⁶⁴ beta-site APP-cleaving enzyme 1 [BACE1] inhibitors for Alzheimer dementia,⁶⁵ proprotein convertase, subtilisin/kexin type 9 [PCSK9] monoclonal antibodies to lower low-density lipoprotein levels⁶⁶) are undergoing advanced clinical trials. Discoveries such as these will directly facilitate the goals of the Precision Medicine Initiative.⁶⁷

Use of other approaches, such as identification of common variants of small effect, might be less effective at facilitating drug development. For example, of 348 proteins specifically targeted by current therapeutics, 42.5% are encoded by a gene responsible for a Mendelian phenotype, whereas only 28.2% of proteins targeted by current therapeutics are encoded by a gene found within GWAS signals (the closest downstream and upstream genes were counted per intergenic signal, and all overlapping genes were counted per coding signal). Accounting for the over-representation of genes underlying Mendelian phenotypes in GWAS signals, 27.3% of proteins targeted are only encoded by a gene underlying a Mendelian phenotype, whereas 13.6% of proteins targeted are found only in a GWAS signal. Moreover, compared to therapeutics that are still in clinical trials, currently approved therapeutics are enriched with drugs that target a protein encoded by a gene in which mutations are responsible for a Mendelian phenotype (32.8% versus 42.5%), suggesting that drugs associated with a gene underlying a Mendelian phenotype more often receive FDA approval. No such relationship is observed for genes found within GWAS signals (28.2% are FDA approved, whereas 29.4% are in clinical trials). Accordingly, using information about whether a target protein is encoded by a gene underlying a Mendelian phenotype might help to stratify drug candidates for development.

Gene-Discovery Efforts for Mendelian Phenotypes to Date

The first successful efforts to identify genes underlying Mendelian phenotypes often required extensive prior knowledge of disease biology, including the identity of the affected protein. In 1986, discovery of mutations causing chronic granulomatous disease in CYBB (MIM: 300481) demonstrated that mapping followed by sequencing of genes within the maximum-likelihood interval offered a promising alternative for discovering genes underlying disease, and during the next 10 years, 42 genes associated with Mendelian phenotypes were identified via positional cloning.⁶⁸ The ensuing two decades witnessed a steady accumulation of genes discovered to underlie Mendelian phenotypes by a combination of positional cloning and candidate-gene approaches. However, it became increasingly obvious that gene identification for most Mendelian phenotypes without a known cause was difficult via these approaches. Gene-discovery strategies based on WES and WGS introduced powerful alternatives that were agnostic to both known biology and mapping data.^{19,22,23,69–72} Combined with conventional genetic approaches, WES and WGS have proved to be disruptive technologies that have rapidly accelerated the pace of discovery of genes underlying Mendelian phenotypes, such that the pace of gene discovery has increased from an average of ∼166 per year between 2005 and 2009 to 236 per year between 2010 and 2014. Between January of 2010 and February 2015, ∼555 and ∼613 genes associated with monogenic Mendelian phenotypes were discovered via next-generation sequencing approaches and conventional approaches, respectively. However, over this time period, there has been a rapid shift toward increasing the use of WES and WGS, and since 2013, WES and WGS have made almost three times as many discoveries as conventional methods (Figure 4; Supplemental Material and Methods; Figure S2).

Figure 4.

Approximate Number of Gene Discoveries Made by WES and WGS versus Conventional Approaches since 2010

Since the introduction of WES and WGS in 2010, the pace of discovery of genes implicated in Mendelian phenotypes per year has increased substantially, and the proportion of discoveries made by WES or WGS (blue) versus conventional approaches (red) has steadily increased (see Supplemental Material and Methods for a detailed description of the analysis). Since 2013, WES and WGS have discovered nearly three times as many genes as conventional approaches.

Although substantial progress has been made toward identifying the genetic basis of Mendelian phenotypes, the genes underlying about half of all known Mendelian phenotypes (i.e., 3,152) have not yet been discovered, despite the fact that ∼20% (i.e., 643) have been mapped (∼80% with robust linkage data [e.g., significant linkage to a single region or recurrent structural variants (SVs) involving the same region] according to a manual review) as per data from OMIM (February 2015). Most of these “unsolved” Mendelian phenotypes are rare and often have high locus heterogeneity and/or are intractable to mapping-based approaches because they are caused by de novo mutations in the germline or mosaicism in somatic tissues.^73–75 Sequencing technologies and analytical approaches have now sufficiently matured to make gene discovery at scale for all Mendelian phenotypes feasible and cost effective. To this end, national and international efforts led by the human genetics community have emerged to identify the genetic basis of Mendelian phenotypes at scale even as the number of recognized phenotypes continues to increase each year.^76–78

The Centers for Mendelian Genomics

Widespread, convenient, and cost-effective application of WES and WGS for finding genes underlying Mendelian phenotypes posed a number of challenges when the strategy was first introduced.⁷⁹ Moreover, achieving the goal of finding all genes underlying all Mendelian phenotypes requires searching the entire human population and therefore necessitates a worldwide collaboration among clinicians and scientists to identify and characterize both novel and well-known Mendelian phenotypes. Accordingly, multiple national efforts were initiated to establish the collaborative framework and physical infrastructure necessary for undertaking large-scale identification of affected individuals, genomic sequencing, and gene discovery for Mendelian phenotypes; these included the NHGRI- and NHLBI-supported Centers for Mendelian Genomics (CMGs),⁷⁷ Finding of Rare Disease Genes (FORGE) Canada,⁷⁶ and the Wellcome Trust Deciphering Developmental Disabilities (WTDDD),⁷⁸ each of which was established in 2011. The CMGs consist of three centers: (1) the University of Washington Center for Mendelian Genomics, (2) the Baylor-Hopkins Center for Mendelian Genomics, and (3) the Yale Center for Mendelian Genomics. All of these consortia, as well as a myriad of individual investigators and small research groups, have made major contributions to gene discovery for Mendelian phenotypes over the past 4 years.

The CMGs were expected to make substantial progress toward discovering the genomic basis of most, if not all, known Mendelian phenotypes. Specifically, the CMGs had the following goals: (1) assess the genetic basis of ∼1,000 Mendelian phenotypes in collaboration with investigators worldwide; (2) develop new methods and approaches for discovering the genetic basis of Mendelian phenotypes; (3) generate public resources that can be leveraged by the biomedical community to facilitate investigator-initiated gene-discovery efforts, studies of gene function, and clinical translation and interpretation of human genome variation; and (4) lead and coordinate US efforts with other large-scale projects aimed at discovering genes implicated in Mendelian phenotypes. Key to accomplishing these goals was that collaborating clinicians and investigators were able to access WES, WGS, and technical expertise from the CMGs at no cost and preserve their control over data sharing, analysis, and rights to publish. It was also anticipated that the overall genetic architecture of Mendelian phenotypes would be further elucidated and that novel underlying genetic mechanisms might be revealed.

CMG Discoveries

As of January 2015, 18,863 samples representing 579 known and 470 novel Mendelian phenotypes from 8,838 families (see the CMG investigated phenotypes in the Web Resources) have been assessed by the CMGs in partnership with 529 investigators from 261 institutions in 36 countries (i.e., ∼1 of every 5 countries in the world) (Figure 5). 60% of countries, 32% of institutions, and 20% of investigators are located outside of North America, Europe, or Australia (Figure 6). Exome and whole-genome data have been produced for 16,226 and 96 samples, respectively, and about half of these sequences can be deposited in dbGaP (Figure 5). Additionally, data for newly identified causal variants have been made available through ClinVar and via a new track on the UCSC Genome Browser. Finally, web-based tools developed by the CMGs, such as GeneMatcher and Geno₂MP, provide a mechanism for investigators with a candidate gene for a Mendelian phenotype to connect with other clinicians and/or basic scientists around the world with an interest in the same gene and to link phenotypic profiles to rare variants, respectively.⁸⁰ Accordingly, the CMGs have empowered the entire international rare-disease research community.

Figure 5.

Overview of Deliverables from the CMGs

Collectively, the CMGs have worked with 529 investigators from 36 countries to collect and sequence 16,226 exomes and 96 genomes. Analyses of these data have resulted in 956 discoveries. These discoveries, as well as tools and technical methods developed by the CMGs, have led to the publication of 146 manuscripts.

Figure 6.

Worldwide Interactions with the CMGs

In collaboration with 529 investigators representing 261 institutions in 36 countries (or 1 of every 5 countries [orange] in the world), the CMGs have collected 18,863 samples from 8,838 families. Approximately 60% (n = 20) of these countries are located outside of North America, Europe, or Australia.

To assess progress toward identifying the genes underlying Mendelian phenotypes, it is critical to apply objective discovery metrics. To date, it has been challenging to quantify and compare reported discovery rates across different contexts (e.g., clinical service versus research). Because of its perceived simplicity, one discovery metric that has been suggested is the “solve rate,” or the proportion of investigated families in whom a causal variant for a Mendelian phenotype is identified. This definition is not particularly useful on its own given that one could, for example, achieve a high solve rate by sequencing only families affected by disorders for which the mutated gene was previously known.

In an attempt to provide clearly defined measures, we developed three complementary discovery metrics and applied them to phenotypes studied by the CMGs on the basis of strict criteria for (1) variant causality, (2) definitions of novel and known phenotypes, and (3) definitions of novel and known genes underlying a Mendelian phenotype (Table 1; Figure 7). This was necessary because although multiple guidelines for assessing causality have previously been proposed,^81,82 none have been operationalized, much less used for assessing large-scale gene-discovery efforts studying thousands of families and samples across hundreds of rare Mendelian phenotypes.

Table 1.

Definitions of Terms Used to Characterize Discovery Type

Term	Definition
Phenotype	the collection of observable or measurable traits of an individual
Known phenotype	a Mendelian phenotype with a MIM number
Explained, known phenotype	a Mendelian phenotype with a MIM number and for which a causal variant(s) in one or more genes is known
Unexplained, known phenotype	a Mendelian phenotype with a MIM number and for which no causal variant(s) has been reported
New phenotype	a Mendelian phenotype without a MIM number (MIM number assigned thereafter)
Known gene	a gene in which a causal variant(s) has been previously associated with a Mendelian phenotype
Novel gene	a gene in which a causal variant(s) has not been previously associated with a Mendelian phenotype
Known gene; explained, known phenotype	a Mendelian phenotype with a MIM number and for which a causal variant was found in a gene previously associated with the same phenotype
Novel gene; unexplained, known phenotype	a Mendelian phenotype with a MIM number, for which no causal variant(s) has been reported, and for which a causal variant was discovered in novel gene
Novel gene; new phenotype	a Mendelian phenotype without a MIM number and for which a causal variant(s) was found in a gene in which a causal variant(s) has not been previously associated with a Mendelian phenotype (MIM number assigned thereafter)
Known gene; unexplained, known phenotype	a Mendelian phenotype with a MIM number, for which a causal variant(s) has not been reported, and for which a causal variant was found in a gene previously associated with a different phenotype
Known gene; new phenotype	a Mendelian phenotype without a MIM number and for which a causal variant was found in a gene previously associated with a different phenotype (MIM number assigned thereafter)
Phenotype expansion	expansion of the spectrum of clinical characteristics of an explained, known Mendelian phenotype

Figure 7.

Criteria for Establishing Causality of Discoveries

Flow diagram of decisions and criteria used for establishing whether gene discoveries by CMGs (Table 2) were considered causal by conservative or suggestive guidelines.

The overall diagnostic rate, defined as the proportion of families in whom a causal variant was identified, was 0.31 and 0.40 per conservative and suggestive causality criteria, respectively. This is comparable to diagnostic rates achieved by clinical WES, but neither the diagnostic rate in the CMGs nor its comparison to diagnostic rates of clinical service labs is a highly informative metric of success. On the one hand, families studied by the CMGs are specifically selected to have phenotypes that are less likely to be explained by an already known gene, thereby potentially lowering the CMG diagnostic rate. On the other hand, the CMGs often have the advantage of studying multiple individuals in a family and multiple families affected by the same unexplained phenotype, which is predicted to improve the diagnostic rate.

To date, 647 and 309 genes by conservative and suggestive causality criteria, respectively, or a total of 956 genes, were discovered by the CMGs to be implicated in a Mendelian phenotype (Table 2). Of the genes discovered by conservative criteria, 327 were (1) a gene that was not previously known to underlie a Mendelian phenotype (i.e., novel gene) but was found to be implicated in a known but unexplained phenotype (i.e., a phenotype with an OMIM number but for which no underlying gene was known; n = 25) or a novel phenotype (i.e., without an OMIM number; n = 107); (2) a gene that was previously known to underlie a Mendelian phenotype (i.e., known gene) and explained either a different known (n = 4) or a novel (n = 17) phenotype; or (3) a gene that was previously implicated in a Mendelian phenotype and was now discovered to be associated with an expanded set of clinical features (i.e., phenotypic expansion, n = 174).

Table 2.

Summary of Discoveries of Genes Underlying Mendelian Phenotypes

Discovery Type		Evidence of Causality
Conservative	Suggestive	Total
Known	known gene; explained, known phenotype	320	19	339
Novel	phenotype expansion	174	24	198
known gene; unexplained, known phenotype	4	0	4
known gene; new phenotype	17	7	24
novel gene; unexplained, known phenotype	25	27	52
novel gene; new phenotype	107	232	339
Total novel		327	290	617
Total number of discoveries		647	309	956

Of our gene discoveries, 320 involved identification of a known gene that explained a known phenotype; the vast majority of these phenotypes (e.g., non-syndromic hearing loss [MIM: PS220290, PS124900]), asphyxiating thoracic dysplasia (MIM: PS208500), and oculocutaneous albinism (MIM: PS203100) had high locus heterogeneity. Less commonly, clinical screening failed to identify a causal variant that was discovered by WES, or a family was recognized in retrospect to be affected by an explained, known Mendelian phenotype with a clinical presentation that was unusual but not different enough to be classified as a phenotypic expansion. Overall, the causal-gene-identification rate, defined as the ratio of causal genes identified to Mendelian phenotypes studied, was 0.51 genes identified per Mendelian phenotype studied.

If gene discoveries meeting conservative and suggestive criteria are combined, 617 were (1) a novel gene that was found to underlie a known, unexplained Mendelian phenotype (n = 52) or a novel phenotype (n = 339); (2) a known gene that explained a different known Mendelian phenotype (n = 4) or a novel Mendelian phenotype (n = 24); or (3) a gene underlying a phenotypic expansion (n = 198) (Figure 8). 339 discoveries were for a gene previously known to underlie the Mendelian phenotype studied. Accordingly, the causal-gene-identification rate combining conservative and suggestive criteria was 0.76 genes per Mendelian phenotype studied.

Figure 8.

Breakdown of Discoveries Made in the 1,049 Mendelian Phenotypes Assessed in the CMG Pipeline

Phenotypes entering the CMG pipelines are putatively either new phenotypes or unexplained, known phenotypes. A substantial fraction (i.e., 32%) of phenotypes were found to have causal variants in known genes, consistent with explained, known phenotypes. However, a larger fraction (40%) of phenotypes assessed resulted in discoveries of novel genes in addition to the expansion of 198 Mendelian phenotypes. For ∼28% of phenotypes assessed, no causal variant has yet been discovered. Novel genes are those that were not associated with any Mendelian phenotype when a project was accepted by the CMGs. Phenotypes are defined on a gene- and/or genotype-centric basis—if a novel gene was discovered for a known, explained phenotype, the phenotype was reclassified as a novel phenotype because it is almost certain that deeper phenotyping would reveal (molecular, biochemical, or physiological) differences that distinguish the novel phenotype from the previously known, explained phenotype caused by mutations in another gene.

Analysis of gene-identification rates by mode of inheritance used for modeling segregation in the analysis of each phenotype provides further resolution about the types of Mendelian phenotypes for which gene discovery was successful and the challenges that remain. Gene-identification rates based on conservative criteria ranged from 0.29 (multiple models) to 0.66 (autosomal recessive) (Figure 9A); for comparison, if a causal gene were identified for every phenotype, this ratio would approach a maximum value of 1. Gene discoveries in consanguineous families were sometimes complicated by locus heterogeneity and by the rarity of the phenotype, consistent with a lower-than-anticipated gene-identification rate (0.60).^83,84 Lastly, the novel-discovery rate, or the proportion of Mendelian phenotypes in which the gene was newly discovered to underlie a novel or unexplained phenotype, including phenotype expansions, was 0.52 (Figures 9B and 9C). Thus, the novel-discovery rate based on conservative and suggestive causality criteria was 0.66.

Figure 9.

Discovery Metrics under Different Models of Inheritance for Mendelian Phenotypes Studied by the CMGs

(A) The percentage of Mendelian phenotypes for which a gene was discovered on the basis of conservative causality criteria per different models of inheritance with mapping data (dark green) or without mapping data (light green) is shown. Also shown is the percentage of Mendelian phenotypes for which a causal gene was not found per different models of inheritance with mapping data (dark gray) or without mapping data (light gray). Note that for most phenotypes analyzed under an autosomal-recessive homozygous model that failed, mapping data were available; however, the statistical significance of the mapping data varied (e.g., number and length of runs of homozygosity, magnitude of LOD score, etc.). The mean number of genes discovered per Mendelian phenotype was 0.52 or 0.76 on the basis of only conservative or combined conservative and suggestive criteria, respectively. These figures do not include results from persons found to have more than one Mendelian phenotype.

(B) Classification of discoveries of genes underlying Mendelian phenotypes as known (white squares) or novel (blue squares).

(C) Percentage of Mendelian phenotypes for which a novel discovery (dark blue) or known discovery (light blue) was made on the basis of conservative causality criteria per different models of inheritance. The mean number of novel discoveries per Mendelian phenotype was 0.52 or 0.66 on the basis of only conservative or combined conservative and suggestive criteria, respectively.

Abbreviations are as follows: AD, autosomal dominant; AR, autosomal recessive (when recessive inheritance was clear, but analysis of both consanguineous and non-consanguineous families contributed to the discovery); AR homozygous, autosomal recessive in a consanguineous family; AR heterozygous, autosomal recessive in a non-consanguineous family (i.e., compound-heterozygous mutations).

The criteria that we used to estimate novel-discovery rates were conservative, in large part because causal variants in the same gene had to be identified in two or more families affected by the same Mendelian phenotype or, if a putatively causal variant was found in only one family, because both high-confidence mapping and compelling functional data (e.g., recapitulation of the phenotype in an animal model) were required.⁸¹ Given that many of the Mendelian phenotypes studied are quite rare and that additional families were not available for study, the net effect of imposing conservative causality criteria was a 1.5-fold reduction in the number of Mendelian phenotypes we considered unequivocally explained. Yet, as more families are studied and more data are shared among researchers and clinicians, our results based on suggestive causality criteria are also likely to lead to genuine gene discoveries for Mendelian phenotypes.

Lessons Learned and Challenges

The overall rate of discovery made by the CMGs over the past 3 years has been approximately three discoveries per week, including the identification of Mendelian phenotypes associated with 375 genes (128 by conservative criteria) not previously known to influence human health. This rate shows no signs of decreasing. Discovery at this scale has also provided guidance about the most effective analytical strategies for gene discovery and, in conjunction with the efforts of other research groups, has exposed genetic mechanisms of disease (e.g., de novo mutation, CNVs, digenic inheritance, somatic mosaicism, and variants that are in the same gene but cause both dominant and recessive conditions) that are more common than previously appreciated. For example, de novo mutations were found to be responsible for a wide range of multiple-malformation syndromes, including visceral myopathy⁸⁵ (MIM: 155310), acromelic frontonasal dysostosis⁸⁶ (MIM: 603671), and Gordon syndrome⁸⁷ (MIM: 114300), as well as isolated congenital heart defects^88,89 and diaphragmatic hernia.⁹⁰ Facioscapulohumeral muscular dystrophy 2 (MIM: 158901) segregates as a Mendelian condition but was found to result from digenic inheritance of a D4Z4 microsatellite repeat on a chromosome 4 haplotype permissive for expression of the embedded gene, double homeobox 4 (DUX4 [MIM: 606009]), and a variant in structural maintenance of chromosomes flexible hinge domain containing 1 (SMCHD1 [MIM: 614982]).⁹¹ Mutations in mab-21-like 2 (C. elegans) (MAB21L2 [MIM: 604357]) underlie a spectrum of major eye malformations that segregate as either a dominant or recessive trait.⁹² Similarly, mutations in sodium leak channel, nonselective (NALCN [MIM: 611549]), a gene whose mutations are known to cause an autosomal-recessive condition—infantile hypotonia with psychomotor retardation and characteristic facies⁹³ (MIM: 615419)—were found to also cause a novel autosomal-dominant disorder, CLIFAHDD (congenital contractures of the limbs and face, hypotonia, and global developmental delay) syndrome.⁹⁴

When multiple modes of inheritance are consistent with the segregation pattern observed in a pedigree or there is otherwise uncertainty about the correct mode of inheritance for a phenotype (i.e., requiring that each model be tested separately), it is clear that the rate of gene discovery is considerably lower than when the mode of inheritance is known or easily predicted (Figures 9A and 9C). The application of multiple models was often required because there were too few families for establishing segregation patterns definitively, too little information for establishing the affectation status of an individual or stratifying persons by phenotypic similarity, or a combination thereof. This underscores the need for ascertainment and deep phenotyping of families along with an infrastructure for warehousing and sharing exome and genome data from “mutation-negative” research subjects and clinical cases. Otherwise, the rate of successful gene identification did not differ markedly across different inheritance models.

Our results on novel-discovery rates show that Mendelian phenotypes caused at least in part by de novo mutations are a rich source of novel discoveries. Additionally, because several affected persons with de novo mutations in the same gene can frequently be identified, it is often easier to satisfy conservative causality criteria in these circumstances than in rare recessive conditions. Correspondingly, our results indicate that the rate of novel discovery for Mendelian phenotypes was unexpectedly⁸³ more modest in consanguineous families when it was based on an autosomal-recessive model and conservative causality criteria. However, the rate of novel discovery in consanguineous families varied among centers. The majority of consanguineous families in the CMG dataset are from geographic regions and populations poorly represented in public variant databases, and the ability to deeply phenotype affected individuals is often limited, both of which reduce the effectiveness of excluding non-causal variants.⁸⁴ Additionally, many of the phenotypes assessed under this model in the CMG dataset (e.g., non-syndromic hearing loss, nephrotic syndrome, brain malformations, or intellectual disability) were categorical phenotypes with high locus heterogeneity, which we have demonstrated is associated with lower clinical diagnostic rates (Figure 1). Accordingly, of the 214 discoveries meeting conservative criteria and made on the basis of an autosomal-recessive model, 129 (60%) involved a known gene for an explained, known phenotype, and only 85 (40%) qualified as novel discoveries (Figure 9C).

Phenotypic expansions and known genes associated with a known, unexplained phenotype or a novel phenotype composed a substantial fraction of CMG discoveries. Until recently, the prevailing method by which new Mendelian phenotypes were described relied on circular logic restricted to the features of the first clinical report. This strategy of ascertaining persons with highly similar phenotypes and subsequent gene discovery within the ascertained group biases against uncovering mutations that are in the same gene and that result in new or different sets of phenotypic features. This phenomenon is apparent in the CMG discoveries and is most likely far more common than generally appreciated. In an analysis of data from OMIM (February 2015), 706 genes or ∼24% of all genes associated with a Mendelian phenotype are responsible for at least two “clinically discrete” phenotypes.

No gene with causal variants could be identified (i.e., either no compelling candidate variants could be identified, or the number of candidate variants remained large) for approximately half of Mendelian phenotypes assessed by the CMGs. Failure to identify a gene underlying these phenotypes is most likely explained by a combination of biological, technical, and analytical factors.⁹⁵ WES cannot detect or has a hard time detecting many types of variants (e.g., indels, non-coding variants, CNVs, repeat expansions, SVs such as inversions and translocations, etc.). WGS is better at detecting these variants and will be increasingly utilized as the cost of this technology is reduced. In ongoing pilot testing, the CMGs conducted WGS on ∼100 samples selected for a variety of reasons, including insufficient DNA quantity for WES, suspicion that the causal variant would be unlikely to be detected by WES, or failure to identify a plausible candidate gene by prior WES. To date, this pilot testing has not led to a discovery that could not be detected or interpreted by WES or genotyping arrays (e.g., for CNVs). We anticipate that this will change as more Mendelian phenotypes fail discovery efforts using WES and are re-examined by WGS. However the lack of WGS discoveries in our limited pilot is consistent with a recent study that found that only ∼15% of causal variants identified by clinical WGS would most likely have been missed by WES. Moreover, although significant, this percentage is not high enough to justify the current additional costs of using WGS on a large scale for gene discovery, especially because the same study found that WGS applied to dominantly inherited disorders usually identified a large number of candidate variants for which pathogenicity could not be determined.⁹⁶ As a result, increasing the use of WGS will require expanded informatics infrastructure and greatly improved tools and methods for establishing causality at scale among the consequentially inflated catalog of candidate variants (e.g., synonymous variants, non-coding variants, SVs, etc.).^81,97,98

Impact of CMGs Accomplishments on Clinical Care and the Scientific Community

The translation and impact of the CMGs’ efforts on clinical care has been immediate and substantial.⁹⁹ Approximately 521 novel Mendelian phenotypes have been delineated, and the clinical features of 219 known Mendelian phenotypes have been expanded. The net effect of these advances will be a better understanding of the genetic and allelic architecture of Mendelian phenotypes, an ability to distinguish among similar phenotypes, and improved ability to facilitate molecular diagnosis for disease. These results are already having substantial clinical implications. For example, variants in genes identified as underlying Mendelian phenotypes since 2011 by the CMGs and other members of the Mendelian genetics community represent ∼30% of positive findings from clinical WES.³⁶ This result emphasizes the ongoing importance of discovering genes underlying disease. Moreover, CMG discoveries have added several hundred putative starting points for the development of targeted therapeutics for Mendelian phenotypes.

In aggregate, CMG-facilitated resources and discoveries have resulted in 146 publications over the past 2.5 years (Table S3). The discordance between the number of discoveries made and the number of publications to date reflects, in part, the time required after gene discovery for identifying and testing additional families and elucidating the underlying molecular, developmental, and physiological mechanisms of disease. These efforts are undertaken foremost to ensure the validity of the discovery but also to increase the likelihood of publication in a high-impact journal.^81,82 Accordingly, most of the discoveries made by the CMGs have yet to be reported because they were made recently, and additional experimentation is underway. Nevertheless, this delay is of considerable concern given the value of gene discovery as a stimulus for basic research and improved clinical care. Alternative methods of rapidly reporting gene discoveries (e.g., online reporting in centralized databases, rapid communication in print journals, etc.) have been proposed^100,101 but face barriers to implementation, including the cost of developing and maintaining an online discovery warehouse, a lack of consensus as to how to safeguard investigator attribution and provide credit for academic advancement, and the benefits investigators perceive to keeping such information privileged.

Beyond discovering genes implicated in disease, the CMGs have developed and disseminated both methods and tools to (1) enable large-scale inventory and indexing of Mendelian phenotypes, as well as provide an analysis module that enables rapid and flexible filtering of WES and WGS data coupled with the family or cohort phenotype data (PhenoDB),^80,102(2) facilitate candidate-gene comparisons among investigators (GeneMatcher),⁸⁰ (3) reconstruct extended pedigree relationships from genotype data (PRIMUS [Pedigree Reconstruction and Identification of a Maximally Unrelated Set]),¹⁰³ (4) perform quality control of variant data (VAT),¹⁰⁴ and (5) make computational predictions of the impact of sequence variants, including non-coding regions (ALOFT [Annotation of Loss-of-Function Transcripts], CADD [Combined Annotation-Dependent Depletion],⁹⁸ and FunSeq¹⁰⁵) (Figure 5).

Opportunities

Achieving NHGRI’s goal of understanding the genetic basis of inherited disease will ultimately require discovering the gene underlying virtually every Mendelian phenotype. This achievement will be a cornerstone of precision medicine.⁶⁷ Today, and for the foreseeable future, the vast majority of genetic diagnostic tests, primary and incidental results returned to families, and results that guide clinical management in children and adults are based on discoveries of genes underlying Mendelian phenotypes. The rapidly growing catalog of phenotypes and causative variants for Mendelian phenotypes immediately (1) makes it possible for clinical laboratories to rapidly improve rates of molecular diagnosis for rare diseases, (2) advances the development and re-purposing of therapeutics, and (3) empowers clinicians to improve the care and management of individuals with a very wide spectrum of conditions. Accordingly, discovering the cause of most, if not all, Mendelian phenotypes could transform the practice of medicine and the care of families.

Of broader interest to the biomedical research community, discovery of genes implicated in all Mendelian phenotypes will (1) connect genes and their protein products to biological systems and clinical phenotypes, (2) provide a pathway for developing better strategies to find genetic and environmental modifiers underlying common diseases, and (3) enable understanding of the functional and phenotypic consequences of non-coding variation. Indeed, it is becoming increasingly clear that, in the absence of deep phenotypic characterization, genomic information even from a large number of individuals is of limited value. Moreover, what can be understood about gene function and phenotypic consequences is proportional to the breadth and diversity of conditions studied. To this end, our results and experiences suggest that the full power of human genomics to explain fundamental biological processes and rapidly transform medical care might be more readily realized by discovery of the genetic basis of all Mendelian phenotypes rather than large-scale sequencing of a handful of common diseases.

During the next 5 years, there will be a growing need for (1) keystone projects (e.g., to delineate and deeply phenotype all Mendelian phenotypes, to catalog all validated large-effect variants underlying Mendelian phenotypes, etc.) to empower the biomedical community and (2) large-scale projects to explain the functional consequences of variants responsible for all Mendelian phenotypes. These projects will generate a complete catalog of the phenotypic characteristics for all Mendelian disorders and the relationships that distinguish them; identify, index, and warehouse all causal variants linked to all Mendelian phenotypes and the range of different phenotypic features observed within each individual Mendelian disorder; and pioneer new approaches and analytical methods that can be exploited at scale for the discovery of genes underlying Mendelian phenotypes and rare disorders caused by non-traditional modes of inheritance (e.g., oligogenic, de novo, somatic mosaicism, uniparental disomy, etc.) and for explaining their overall contribution to the burden of human disease.

Web Resources

The URLs for data presented herein are as follows:

• ALOFT, https://www.github.com/gersteinlab/aloft

• Centers for Mendelian Genomics, http://mendelian.org/

• Centers for Mendelian Genomics investigated phenotypes, http://mendelian.org/investigated-disorders

• ClinVar, http://www.ncbi.nlm.nih.gov/clinvar/

• dbGAP, http://www.ncbi.nlm.nih.gov/gap

• ENCODE, http://www.genome.gov/encode/

• Ensembl, http://www.ensembl.org/index.html

• GeneMatcher, https://genematcher.org/

• GeneReviews, http://www.ncbi.nlm.nih.gov/books/NBK1116/

• Geno₂MP, http://geno2mp.gs.washington.edu/

• NHGRI GWAS Catalog, http://www.genome.gov/gwastudies/

• OMIM, http://omim.org

• PhenoDB, http://phenodb.net

• PRIMUS, https://primus.gs.washington.edu/primusweb/

• Therapeutic Target Database, http://bidd.nus.edu.sg/group/ttd/ttd.asp

• UCSC Genome Browser, http://genome.ucsc.edu/