Precision medicine in rare diseases: What is next?

Molecular diagnostics is a cornerstone of modern precision medicine, broadly understood as tailoring an individual's treatment, follow‐up, and care based on molecular data. In rare diseases (RDs), molecular diagnoses reveal valuable information about the cause of symptoms, disease progression, familial risk, and in certain cases, unlock access to targeted therapies. Due to decreasing DNA sequencing costs, genome sequencing (GS) is emerging as the primary method for precision diagnostics in RDs. Several ongoing European initiatives for precision medicine have chosen GS as their method of choice. Recent research supports the role for GS as first‐line genetic investigation in individuals with suspected RD, due to its improved diagnostic yield compared to other methods. Moreover, GS can detect a broad range of genetic aberrations including those in noncoding regions, producing comprehensive data that can be periodically reanalyzed for years to come when further evidence emerges. Indeed, targeted drug development and repurposing of medicines can be accelerated as more individuals with RDs receive a molecular diagnosis. Multidisciplinary teams in which clinical specialists collaborate with geneticists, genomics education of professionals and the public, and dialogue with patient advocacy groups are essential elements for the integration of precision medicine into clinical practice worldwide. It is also paramount that large research projects share genetic data and leverage novel technologies to fully diagnose individuals with RDs. In conclusion, GS increases diagnostic yields and is a crucial step toward precision medicine for RDs. Its clinical implementation will enable better patient management, unlock targeted therapies, and guide the development of innovative treatments.

Molecular diagnostics is a cornerstone of modern precision medicine, broadly understood as tailoring an individual's treatment, follow-up, and care based on molecular data. In rare diseases (RDs), molecular diagnoses reveal valuable information about the cause of symptoms, disease progression, familial risk, and in certain cases, unlock access to targeted therapies. Due to decreasing DNA sequencing costs, genome sequencing (GS) is emerging as the primary method for precision diagnostics in RDs. Several ongoing European initiatives for precision medicine have chosen GS as their method of choice. Recent research supports the role for GS as first-line genetic investigation in individuals with suspected RD, due to its improved diagnostic yield compared to other methods. Moreover, GS can detect a broad range of genetic aberrations including those in noncoding regions, producing comprehensive data that can be periodically reanalyzed for years to come when further evidence emerges. Indeed, targeted drug development and repurposing of medicines can be accelerated as more individuals with RDs receive a molecular diagnosis. Multidisciplinary teams in which clinical specialists collaborate with geneticists, genomics education of professionals and the public, and dialogue with patient advocacy groups are essential elements for the integration of precision medicine into clinical practice worldwide. It is also paramount that large research projects share genetic data and leverage novel technologies to fully diagnose individuals with RDs. In conclusion, GS increases diagnostic yields and is a crucial step toward precision medicine for RDs. Its clini-

Introduction
Rare diseases (RDs) collectively affect 3.5%-5.9% of the global population, which equates to 260-440 million individuals. However, the definition of From the symposium: Cutting-edge implementation of precision medicine in Europe. rare is not uniform across the world [1]. In some countries, the definition is based upon a maximum number of affected individuals in that country, such as in the United States and Japan where a rare disease is one that affects, respectively, fewer than 200,000 and 50,000 people living in the country at a given time [2,3]. In contrast, the countries in the European Union (EU) have defined a RD as one affecting fewer than 1 in 2000 people [4]. Using this definition, RDs affect approximately 30 million people in the EU. This lack of a worldwide consensus is problematic, and a global definition would benefit the field. In a review from 2015, Richter et al. identified 296 definitions from 1109 organizations and suggested that the average of those definitions (1 in 2500 individuals) as a suitable global definition of RDs [5]. Similarly, there is no agreement on the total number of RDs. Available catalogues suggest that around 7000-9000 different RDs exist [6,7]. As increasingly more disease entities are described, the total number of expected RDs might in fact exceed 10,000 [8]. RDs are often lifelong and many times confer significant morbidity, reduced life span, and/or affect reproductive capability [4,9]. For instance, one in four deaths in the neonatal intensive care units is linked to a rare genetic disorder, and patients with RDs have been reported to account for 10% of hospital discharges, also having on average longer hospital stays than the general population [10][11][12]. Thus, there is increasing recognition of RDs as a public health priority.
Traditionally, genetic investigations in individuals with RDs were primarily done to confirm a clinical suspicion of a Mendelian disorder or a chromosomal syndrome, and pathogenic findings were used to inform genetic counseling and to offer carrier testing, prenatal diagnostics, and preimplantation genetic testing to couples with a high recurrence risk. Today, establishing an accurate molecular diagnosis is a cornerstone of precision medicine, broadly defined as tailoring an individual's treatment, follow-up, and care based on genetics. For an affected individual, a genetic diagnosis may result in a deeper understanding of what symptoms to expect, how and when to do surveillance, and for some, targeted therapy.
Even though 40%-72% of RDs are classified as genetic [1,4], only a fraction of the patients are currently correctly diagnosed, whereas many remain undiagnosed or receive an incorrect diagnosis. Even those who eventually receive an etiological diagnosis often face considerable challenges along the way. This process, often referred to as "diagnostic odyssey," can take years and includes meeting with several specialists and undergoing invasive exploratory procedures, as well as numerous laboratory and other diagnostic investigations [13]. There is increasing evidence that a broad use of genome sequencing (GS) as first-line investigation in patients with RDs may shorten the time to diagnosis overall and solve the diagnostic odyssey problem.
This review focuses primarily on GS as a key genomic technology for precision medicine in RDs. We begin by highlighting current initiatives for precision medicine in different European countries. Next, we review the indication for genomic testing, current diagnostic yields, and challenges with the interpretation of GS data. We discuss how precision diagnostics can enable the development of new treatments for RDs and analyze essential requirements for the implementation of precision medicine into clinical practice. Finally, we explore novel technologies and important initiatives to diagnose all RDs.

Genomics England and the NHS genomic medicine service
In 2012, at the London Olympics, former United Kingdom's (UK) prime minister David Cameron announced the 100,000 Genomes Project and expressed the UK's long-term ambition to implement GS into health care. To run this projectinitially focused on RDs, cancer, and infectious diseases-genomics England was established by the UK Department of Health. The 100,000 Genomes Project was the first major national genomic initiative that succeeded in recruiting patients through routine care, sequencing genomes at large scale, and returning validated results through the national health system (NHS) to participants. In the field of RDs, GS within the 100,000 Genomes Project led to a genetic diagnosis in 25% of participants [14]. To further integrate GS into routine health care and as part of an overall approach to standardize genomic testing and embed genomic medicine into end-to-end pathways of care, the NHS genomic medicine service (NHS GMS) was launched in October 2018. The NHS GMS includes a network of seven Genomic Laboratory Hubs that, together with the NHS GMS alliances, guarantee countrywide coverage and equal access to genomics services in the UK. To streamline genomics testing, a big effort has been put into identifying eligibility criteria for genetic tests and suitable first-line investigations for different disorders together, as well as identifying which specialists would be expected to request those tests. This work has led to the creation of the National Genomic Test Directory, which is reviewed and updated annually and is expected to have a significant impact, potentially becoming an international standard [15].

Genomic Medicine Sweden-rare diseases (GMS-RD)
In Sweden, the Genomic Medicine Sweden (GMS) initiative, launched in 2017, is constructing a national infrastructure for implementation of precision medicine [16]. The aim of the GMS working group for RDs (GMS-RD) is to improve diagnostics by implementing genomics-based methods into routine health care. One primary long-term activity is to build a clinical database for genotypephenotype matching in order to facilitate interpretation of genetic results. Presently, clinical GS is first-line diagnostics for patients with RDs in three out of seven Swedish regions, and more than 5000 clinical GS have been performed within Swedish health care in the field of RDs until 2021 [16,17].

TRANSLATE-NAMSE
In Germany, the innovation project TRANSLATE-NAMSE-designed to improve the care of patients with RDs-was active from April 2017 until September 2020 [18]. Ten German centers recruited >5000 patients who had not received a diagnosis from standard of care, including both children (64%) and adults (36%). To date, exome sequencing (ES) has been done for 1599 individuals, resulting in 506 confirmed diagnoses (32%; 415 < 18 years; 91 ≥ 18 years) [19]. Further, national clinical workflows for diagnosis and care of RDs were generated [20]. The project also invested in improving the transition from pediatrics to adult care and in creating clinical pathways for selected groups of RDs [21].

Plan
France Médecine Génomique 2025 (PFMG2025), launched in 2017, aims to integrate genomic medicine into health care, promotes research and innovation, and drives economic growth by producing 235,000 genomes annually. The initiative builds on national plans for RDs established since the early 2000s which have been instrumental in building several networks comprising referral and competence centers for specialized medical care. A catalog of medical indications has been compiled, highlighting areas in which GS could offer significantly improved diagnostics, with 60 RDs currently under evaluation for health care-funded GS. PFMG2025 has set up working groups to address the ethical, regulatory, and societal issues of GS as well as to perform health-economics evaluation and provide training. Additionally, a genomic medicine infrastructure with high-throughput sequencing instruments and modern standards for laboratory and data processing has been established.

European Reference Networks (ERNs)
In 2017, the EU launched the European Reference Networks (ERNs) to address the needs of patients with rare and complex diseases. ERNs are crossborder virtual networks of health care providers, expert centers, research institutions, and patient organizations. There are 24 ERNs, each focused on a specific rare disease area-for example, eye diseases and immune disorders-aiming to foster collaboration and information-sharing to improve the quality of life and care for patients with RDs.

The genomic revolution
In 1966, Dr. McKusick published the first edition of his book Mendelian Inheritance in Man, covering 1400 phenotypes. At the time, no human trait had been linked to autosomes, whereas about 60 traits had been identified as X-linked [22]. Studies of rare phenotypes and inventions, such as Sanger sequencing, polymerase chain reaction (PCR), and positional cloning, were instrumental for the tremendous knowledge gain that the field of RDs experienced over the subsequent years. By 2001, when the first draft of the human genome was made available [23,24], 1000 Mendelian genes had been identified [25]. In the early 2000s, diagnostic assays, such as chromosome analysis, PCR, and Sanger-based diagnostics, were routinely available in clinical genetics laboratories. The advent of chromosomal microarray enabled a new understanding of structural and copy number variants in the genome and became a key diagnostic test for children with developmental delay and/or congenital malformations [26]. Shortly thereafter, short-read next-generation sequencing technologies (NGS) started taking off and further sped up the discovery of the molecular basis for many Mendelian diseases [27][28][29]. As of 23rd March 2023, the Online Mendelian Inheritance in Man catalog contains 7352 phenotypes with a known molecular cause, with more than 5000 having been molecularly characterized in the last 20 years [30].
The core strength of NGS based on short reads is the ability to simultaneously interrogate multiple genomic regions, something that quickly becomes very labor intensive with conventional Sanger sequencing. This has made NGSbased DNA-sequencing suitable for the diagnostics of Mendelian RDs characterized by genetic heterogeneity-for example, non-syndromic hearing loss and intellectual disability (ID) [31]. NGSbased DNA-sequencing comes in different "flavors," with the three most common diagnostic applications being targeted panels, ES, and GS. This review will focus on GS-sequencing the entire genome-which, due to the decreasing cost of DNA sequencing, is expected to become the leading method for diagnostics of RDs. The main advantages of GS compared to other targeted NGS approaches (including gene panels and ES) are the possibility to confidently capture single nucleotide variants (SNVs) and several types of structural variants (SVs)-including most balanced rearrangements-as well as short tandem repeats [32][33][34][35][36]. Furthermore, even though ES supposedly captures all the coding regions of the genome, GS is superior to ES even in the detection of coding SNVs, likely due to a more even coverage [37].

Indication for GS in clinical care
Indications for GS are still being discussed [31,38]. The American College of Medical Genetics and Genomics (ACMG) recently recommended that GS or ES should be used as a first-or second-tier test for patients with congenital anomalies and/or ID [39]. The NHS National Genomic Test Directory discussed above is also an important initial step in defining the indications for different genetic testing, including GS. The NHS National Genomic Test Directory is regularly reviewed based on emerging evidence and diagnostic discoveries.
Even so, recommendations for specific clinical indications are largely missing, only available locally or through guidelines issued by scientific societies or expert groups. Overall, GS is suitable as firsttier analysis for most patients with a suspected genetic disorder characterized by genetic heterogeneity as well as all patients for which previous diagnostic tests have failed to provide a molecular explanation (Fig. 1). It is important to consider that for most patients eligible for ES, GS is also appropriate. However, special consideration is needed for some subcategories, such as patients with suspected mosaic disorders. In such cases, deep sequencing of relevant genes through targeted panels or ES is preferred because the higher sequencing depth (i.e., number of times a genomic region is sequenced) offered by those applications increases the diagnostic sensitivity. In contrast, short tandem repeats-a common cause of neurological disorders-are better detected by GS than ES, and therefore GS is a better choice for such patients [40]. Of note, all current short-read DNA sequencing methods fail to identify methylation defects and therefore cannot be used in the diagnostics of methylation defects.
Although requesting genetic tests was previously predominantly done by clinical geneticists, today this is increasingly done by other types of medical specialists-for example, pediatricians, oncologists, cardiologists, and neurologists (Internal data, NHS genomic medicine service; Karolinska University Hospital). Clinical geneticists, while continuing to request genetic tests, also have an increasing role in assisting the different specialists in the selection of the appropriate method and testing strategy depending on the clinical question. In fact, ordering a genetic test requires knowledge of advantages and disadvantages with the different methods, as well as deciding if a patient should be sequenced alone-so-called singleton analysisor if family members should be sequenced at the same time to decrease the number of variants to be reviewed (Fig. 1). A common approach is the socalled trio design, in which genomic DNA from the affected probands and their parents are analyzed together. A trio design is especially suitable when both parents are healthy and a de novo disorder is suspected, such as in most children with ID syndromes [28,41].
Although the entire genome is sequenced, a clinical GS is usually analyzed with the help of virtual gene panels (Fig. 1). Virtual gene panels are lists of genes with a proven causality association to a specific disorder or group of disorders that can be used to filter the GS data. Narrowing the number of genes tested reduces the number of variants that need to be reviewed by the laboratory and clinical geneticists. It also reduces the risk for incidental findings. Due to the rarity of some disorders, it is sometimes difficult to achieve consensus on which genes should be included in a gene panel [42]. PanelApp-an online database of gene panels The suggested panels (Undiagnosed metabolic disorders, paediatric disorders, and retinal disorders) refer to existing gene panels in PanelApp. HPO (human phenotype ontology)-based panels can be generated with Phenomizer [44]. MRI, Magnetic resonance imaging.
in which each gene is reviewed by multiple experts worldwide-was developed by genomics England in the UK to address the lack of uniformity between gene panels [43]. PanelApp is currently the chosen source of gene panels for GS and other genomic tests performed within the NHS [15]. Alternatively, personalized virtual gene panels can be created using human phenotype ontology (HPO) terms [44,45] or through recommendations of expert groups or scientific societies [46,47]. The use of Pan-elApp and/or HPO-based patient-specific panels reduced the average number of variants to assess from 45 to 7 in patients who underwent clinical ES, with an overall diagnostic yield of 24%, which shows the effectiveness of virtual gene panels [48].
Nonetheless, the first results from the 100,000 Genomes Project show that 26% (141 out of 535) of molecular diagnoses were made by identifying pathogenic variants in known disease genes that were not part of the original in silico panels. The variants were instead detected through manual review or with the help of tools such as Exomiser, which prioritizes variants based on the patient's HPOs [14,49].

Whole-genome sequencing as a first-tier diagnostic test?
There is a growing body of literature concerning the advantages of GS over standard of care genetic testing, which includes both specific genetic tests based on clinically suspected genetic conditions as well as genome-wide screening tests such as chromosomal microarray and/or ES. Neonatal intensive care units have been a natural setting for early introduction of GS-in particular rapid GS (rGS), a type of GS and bioinformatics pipeline with focus on delivering quick results. Studies have shown that rGS has a two-to-threefold higher diagnostic yield compared to standard of care [50][51][52][53], and a mean saving per person of $100,440 [52]. In the NICU-seq project-a randomized time-delayed clinical trial-the rGS arm in neonatal intensive care units showed not only a twofold higher diagnostic yield but also a twofold higher rate of changes in clinical management compared to the standard of care arm [53]. Interestingly, a doubling of diagnostic yield was seen even in the standard of care arm when the delayed GS results were disclosed after 90 days, confirming the diagnostic power of GS [53]. The NSIGHT2 study compared rapid ES and rGS (n = 94) in critically ill infants with diseases of unknown etiology and demonstrated better analytical performance of rGS, although in this study both methods resulted in similar diagnostic yields [54]. Extended analysis with the trio approach in patients with negative results increased diagnostic yield by <1%, suggesting that singleton rGS may be the most effective approach in neonatal intensive care units [54]. Confirming the clinical utility of rGS, participating clinicians rated the investigation useful in the vast majority of patients, with changes in management being more likely when results were positive and turnaround time was shorter [55]. The NHS GMS has introduced rGS for neonatal and pediatric ICU patients as a national provision with a turnaround time of less than 10 days and a diagnostic yield of over 40%, with a clear impact on clinical management and outcomes.
Especially suitable for broad investigation such as ES and GS are genetic heterogenous conditions, such as ID and neurodevelopmental disorders (NDD). In 2014, Gilissen et al. were the first to apply GS to study individuals with yet unexplained ID, with good results [56]. Subsequent studies confirmed a diagnostic superiority of GS in patients with ID and NDD, not only because of higher diagnostic yields but also because it shortens the time to diagnosis compared to when combinations of different tests are used-for example, FMR1 CGG repeat expansion test, chromosomal microarray, and/or ES [57][58][59]. Moreover, a multistep approach-in which GS is only used as a second-or third-line analysis-also carries the risk that no further investigations are pursued despite a negative result in the initial investigations. This will leave many individuals without a molecular diagnosis that could have been achieved were GS used as a first-line investigation [57]. Such observations are consistent across a broad spectrum of pediatric disorders with suspected genetic causes [32,[60][61][62][63][64] and hold true even in cohorts with a high degree of adult patients [14,65,66]. In the 100,000 Genomes Pilot on rare-disease diagnosis, a 25% overall diagnostic yield was reported by applying singleton or trio GS on 2183 probands, of which 74% were adults [14]. Some will argue that certain distinct presentations in adults (e.g., hereditary breast cancer) may still warrant investigation by specific targeted panels. Once end-toend sequencing prices decrease even further, GS may be used even for those indications due to more comprehensive variant calling.

Clinical interpretation of GS data
The process of clinical genome analysis, as illustrated in Fig. 2, encompasses multiple steps from sequencing to bioinformatic analysis and clinical interpretation, each of which require rigorous quality control [66,67]. It is important that only variants with a high probability of being causative are reported back to health care providers and patients. The ACMG classification criteria have been widely accepted as the standard for variant classification [68][69][70]. Gene-specific adjustments may be necessary and have mainly been developed as part of the ClinGen initiative [71][72][73][74]. Guidelines for interpreting noncoding variants are emerging [75]. Only variants classified as pathogenic or likely pathogenic are relevant to clinical decisionmaking, whereas reporting of variants with uncertain significance (VUS) should be done with caution. Nonetheless, depending on laboratory policies, VUS may sometimes be communicated to the referring clinicians and to patients. We recommend reporting of VUS only when relatively simple additional analyses-such as segregation in large families for dominant disorders, or parental analysis in suspected de novo disorders-could enable a reclassification to either likely pathogenic or likely benign (Fig. 2). It is important to note that the available evidence for a specific variant might change over time. Therefore, regular reevaluation of variant classification is recommended [76,77]. In addition, multiple diagnoses may be detected in some individuals with suspected RDs. Posey et al. demonstrated that out of 2076 individuals

Fig. 2 Simplified genome sequencing (GS) workflow and expected diagnostic outcomes. Bioinformatics analysis of GS data can detect a broad range of genetic variants, including single nucleotide variants (SNVs), copy number variants (CNVs), structural variants (SVs), short tandem repeats (STRs), and uniparental disomy (UPD). GS data in clinical settings is most
often filtered with the help of virtual gene panels, and rare variants are reviewed for pathogenicity and classified according to American College of Medical Genetics and Genomics (ACMG) guidelines. About 40% of patients will receive a molecular diagnosis, ∼10% might have a "strong" VUS reported, and ∼1% will have an incidental finding/secondary finding reported according to national and local policies. Overall, ∼50% will at first remain undiagnosed and might benefit from RNA-sequencing, periodic GS reanalysis every 2-3 years, and inclusion in research programs. diagnosed using ES, 4.9% had multiple molecular diagnoses explaining distinct parts of the clinical presentation [78]. Finally, laboratories offering genomic testing need to have clear policies for handling secondary findings and incidental findings. However, recommendations on this matter differ widely between countries and organizations. Some argue that pathogenic variants in "actionable" genes should be actively excluded, some recommend the use of virtual panels to decrease the likelihood that they are reported, and others use an opt-in or opt-out approach [79][80][81][82][83][84][85][86].

Periodic reanalyzes increase diagnostic yields
Regular reanalysis of GS data is a powerful way to improve diagnostic yields because novel diseasecausing genes are discovered at a steady state, and bioinformatic and interpretation tools-as well as databases of genetic variations-are continuously improved. The ACMG recommends that such reevaluation and reanalysis of genomic tests be performed every 2 years [87]. Costain et al. showed that reanalysis 2 years after the initial investigations yielded a diagnosis in seven out of 64 (10.9%) affected individuals lacking a molecular diagnosis [88]. One diagnosis was made because additional clinical information was available to guide the GS interpretation, whereas in six affected individuals, the disease-causing genes were discovered only after the first analysis was done. Even though comparable results have been shown for reanalysis of ES data [89,90], GS has the potential to result in more diagnoses than ES over time. This is mainly due to the possibility of detecting a wide range of genetic variants and investigating noncoding regions [60,91]. Even though a clinical analysis of genome data today is typically limited to coding regions, as more disease-causing variants in noncoding regions are discovered, they can be added to the reanalysis in the same way as novel disease-causing genes. Improved understanding of genome function and disease associations over the next few years is likely to result in even higher diagnostic yields after periodic reanalysis of GS data from individuals without diagnosis.

Cross-disciplinary integrated diagnostic units
The growing complexity of modern medicine necessitates closer collaboration between different clinical and laboratory specialties and among other key professionals. Organizational structures that prioritize cross-disciplinary teams where clinical specialists work side-by-side with experts in genomics have a distinct advantage in facilitating timely diagnostics and expediting individualized patient management [92]. Such approaches are particularly valuable for the investigation of acute and severe conditions in which individualized treatments are becoming increasingly available-such as epilepsy, primary immunodeficiencies, and acute liver failure. The field of inborn errors of metabolism has been spearheading the field of personalized medicine since the 1900s because the understanding of underlying mechanisms can often result in specific treatments, such as specific diets or supplements, recombinant enzymes, and small molecule drugs. More recently, antisense/oligonucleotide technology, allogeneic stem-cell transplantation, and gene therapy have emerged as treatment alternatives. Initiation of treatment in early disease stages can often prevent serious handicaps or early death, making rapid diagnostics essential [93]. As such, diagnostics of inborn errors of metabolism that traditionally have relied on highly specialized biochemical investigation are now increasingly being complemented with genetic/genomic investigations. The availability of well-established biochemical tests is essential for the functional validation of VUS and can enable higher diagnostic yields.

From diagnosis to treatment: New possibilities
There is some form of precision treatment available for over 500 RDs. Some treatments are very expensive, such as advanced medical therapeutic agents-for example, gene therapies or other targeted drugs or bone marrow transplants-whereas others are inexpensive, such as nutritional supplements or commonly available drugs [94]. An early diagnosis and appropriate interventions can often minimize harm to patients and reduce unnecessary costs for the health care system.
Orphan drugs are drugs developed specifically for RDs. Unfortunately, due to the small number of patients with each specific condition, the development of such drugs is often costly and has limited commercial appeal for pharmaceutical companies. To improve the attractiveness of orphan drug development, the EU implemented new legislation in 2000 in order to provide incentives to pharmaceutical companies, either in the form of scientific advice or various fee reductions [95]. As of 2020, there were approximately 2200 medicines designated as orphan drugs and over 160 authorized orphan medicines in Europe [96]. In addition, Europe has currently approved six gene therapies as orphan drugs (Table 1).
Due to the recent advancements in genomics, many RDs have been further characterized, thus facilitating the understanding of the underlying physiopathology, and creating unique opportunities for the development of precision medicine. For example, the recently gained knowledge that genetic disorders involving the RAS/PIK3CA/AKT/mTOR pathway share common defects with cancers has allowed cancer drug repositioning [97]. These RDs are caused by either germ line variants (e.g., in tuberous sclerosis disorders) or somatic mosaic mutations (e.g., in PIK3CA-related overgrowth disorders, PROS) in genes of the RAS/PIK3CA/AKT/mTOR pathway. In PROS, mutations occur in tissue during embryonic development and are not inherited. Interestingly, the RAS/PIK3CA/AKT/mTOR pathway is frequently affected by somatic mutations in several cancer types. As the same PIK3CA mutations observed in cancer also cause PROS, there is an opportunity for repositioning cancer drugs for the treatment of PROS. Such efforts have yielded positive results: alpelisib-a PIK3CA inhibitor developed to treat breast cancer with PIK3CA mutations-was recently approved by the US FDA for use in patients with PROS [98][99][100]. Similarly, everolimus-an mTOR inhibitor initially approved for oncology and in solid organ transplantation-has been repurposed for patients with tuberous sclerosis disorders [101]. Other therapeutic options are currently being evaluated for patients with overgrowth or vascular malformation involving other genes, such as AKT1 gain of function mutation, hereditary hemorrhagic telangiectasia, or RASopathies [102][103][104]. A definitive molecular diagnosis is necessary for such targeted treatments, emphasizing the importance of genetics diagnostics for the success of precision medicine.

Education
Despite the rapid pace of technical advancements in genomics, the integration of these advances into routine health care has been slower. Both health care professionals, patients and other stakeholders must realize that medicine is undergoing significant changes. To ensure that precision medicine fulfills its promises and delivers better health outcomes for patients with RDs, a large effort is required to educate and train present and future health care professionals, and to inform and educate the general population [105]. Improving genomics literacy among medical specialties beyond clinical genetics is essential to ensure widespread access to genomics throughout medicine. The fact that several national genomic initiatives include educational work packages is a clear indication that stakeholders recognize the importance of this challenge. For example, the NHS GMS educational initiative GeNotes is a collection of "just in time" explanations and clinical scenarios designed to guide clinicians using genomics in their practice [106].
The systematic incorporation of genomics and precision medicine into the training curriculums of all health care professionals is crucial to ensure that the next generation of health care professionals is trained to manage genetic information on daily basis. At the same time, high-quality continuing medical education and continuing professional development programs are necessary for training existing health professionals who order genomic tests and make decisions about posttest interventions, including knowledge of available targeted treatment or ongoing clinical trials [107,108]. One current initiative is Global Nursing Network for RDs that was founded during a roundtable meeting held in Singapore in March 2023 to enhance the skills of practicing nurses globally in caring for patients with rare and undiagnosed diseases. Furthermore, in Europe, ERNs are a valuable infrastructure for educating professionals about RDs, with webinars and written educational material available within most ERNs.

International organizations to advance knowledge and advocate for RDs
Several international organizations are focused on advancing RD discovery and diagnosis and are working collaboratively across countries to address common challenges and promote datasharing in both research and clinical settings.  [111]. RDI works as a strong common patient voice to raise awareness, support, and empower RDI members and to advocate for RDs as an international policy priority. International collaboration across multiple stakeholders will be required to actualize the full potential of precision medicine for RDs. International collaboration across multiple stakeholders will be required to actualize the full potential of precision medicine for RDs. Collaboration with patient advocate groups can be supportive in prioritization decisions as they can provide valuable input on the needs and preferences of patients with different RDs. Ultimately, such decisions should follow ethical principles while taking into account the specific context and limitations of the individual health care organizations.
What is next?

Beyond GS: Emerging methods
Successfully solving all RDs is an extremely challenging task, and GS alone will not be enough for a subset of individuals. For instance, it is becoming increasingly evident that there is a significant amount of genetic variation that cannot be captured using short-read sequencing alone. Compared to short-read sequencing, long-read GS shows a fivefold increase in the sensitivity for SVs in the size range of 7-1000 bp [112,113]. Longer reads are also necessary to generate highquality genome assemblies and phase variants, and to resolve complex SVs at the nucleotide level [114]. Novel informatic platforms also enable faster turnaround times that are important in the intensive care setting [115].
Complementing GS/ES analysis by sequencing RNA has been shown to increase the diagnostic yield by 7% to 36% compared to GS alone [116][117][118][119][120][121]. So far, RNA sequencing in RD diagnostics has been mainly used to identify aberrant splicing, but pipelines are being developed to also detect genes with aberrant expression, or transcripts exhibiting mono-allelic expression [121,122]. A major problem is getting access to appropriate cells where the gene is expressed. Hence, clinical laboratories are using either blood samples or cultured fibroblasts to query effects of DNA variants on RNA across broad patient groups. As RNA sequencing is added to diagnostic pipelines for RDs, it is important to standardize interpretation of results and how they impact the clinical classification of genetic variants [123].
Several neurodevelopmental syndromes are caused by pathogenic variants in genes controlling DNA methylation and other epigenetic mechanisms [124]. The methylation profile of several neurodevelopmental syndromes has been explored in recent years, and an increasing number of genetic syndromes have been found to have unique genomic DNA methylation patterns in blood (called "episignatures") [125]. This knowledge is currently being utilized to develop tests that can identify specific DNA methylation patterns of selected known syndromes and imprinting disorders. Such DNA methylation classifiers can be used as a further test to assess pathogenicity of VUS, but also to improve diagnostic yields in patients with undiagnosed RDs [126,127].
The current workflows for diagnosing RDs rely heavily on variants that are functionally annotated or previously known. However, only about 30% of the known variants in ClinVar have a definitive and robust classification as pathogenic/benign, and almost half of them are currently classified as VUS [128,129]. These uncertain variants can potentially explain at least 10% of patients who remain undiagnosed after GS. With GS being envisioned as standard of care and the increasing number of unannotated variants being identified in noncoding regions, there is an urgent need to develop "next-generation functional tests" for the comprehensive and systematic evaluation of thousands of variants from GS data. Multiplex assays of variant effects (MAVEs), which encompass strategies for coding and noncoding sequences, could be used to address this shortage [130,131]. Several recent articles review the technical aspect of MAVEs [130][131][132]. In brief, thousands of genetic variants are simultaneously created and tested in an in vitro system [131]. Although this highthroughput approach in principle allows for systematic screening of all possible nucleotide variants within a gene or region of interest, the MAVE approach is currently limited to research settings.

Data-sharing and matchmaking
A central challenge for RD diagnostics is how to efficiently generate evidence for variant causality, and linking for the first time a gene to a disease. Scalable and global approaches to facilitate responsible data-sharing continue to be needed, but some important progress has been made. The concept of genomic matchmaking was developed in 2015 to discover novel disease-gene relationships through the interrogation of otherwise siloed international datasets [133]. There are three types of genomic matchmaking based on how many parties are involved: Two-sided matchmaking refers to two or more parties with interest in the same gene or variant trying to find each other; onesided matchmaking refers to only one-party sending queries to a dataset; zero-sided matchmaking refers to the scenario in which computer algorithms are used to identify potentially matching cases. The most prominent example of success of two-sided matchmaking is the Matchmaker Exchange (MME), which has been a tool used by the global genetic community for over 7 years [133,134]. The number of cases across the MME is now more than 125,000 from over 13,000 contributors in 102 countries and has facilitated more than 600 published discoveries. Several platforms provide information about the existence of a specific variant and associated information in a public-facing (MyGene2, Geno2MP), one-sided (Franklin), or twosided fashion (VariantMatcher) and are working to facilitate a connection with one another using DataConnect [135]. Several groups are using zerosided matchmaking to compute across a cohort to identify genes with predicted damaging variants in patients with overlapping phenotypes (e.g., the Deciphering Developmental Disorders study) [136].
Moving forward, there are opportunities to further facilitate matchmaking effectively at large scale to rapidly improve our understanding of genetic variation and its impact on human development and health. Data-sharing and matchmaking are important components of three large ongoing efforts in RD research-namely, Care4Rare in Canada, Solve-RD in Europe, and Undiagnosed Diseases Network International (UDNI) worldwide [135][136][137]. All these research programs focus on finding new molecular causes to previously unsolved RDs through data mining and reinterpretation of variants even beyond the coding region. Outcomes from these studies are providing valuable insights into the best way forward.

Concluding remarks
GS as first-line investigation can achieve a molecular diagnosis in more individuals with RDs. The implementation of GS in clinical care at large scale requires a massive educational effort and a multidisciplinary approach to integrate multimodal data, including highly specialized clinical, laboratory, and imaging investigations. Variant interpretation will have to become more robust, with help from complementary tests, such as RNA-sequencing, methylation studies, and MAVEs. Refined diagnostics coupled together with orphan drug development are necessary for precision medicine and better health outcomes for all individuals with RDs. Diagnosing all RDs is an extremely challenging task, and success will depend on innovative approaches, data-sharing, monitoring of outcomes linked to precision therapeutics, and tight collaboration with patient advocacy groups.

Conflict of interest statement
Professor Anna Lindstrand has received honoraria from Illumina. Dr. Guillaume Canaud has the following disclosures: (1)