Implementing genomic medicine in clinical practice for adults with undiagnosed rare diseases

Introduction

Rare diseases are uncommon individually but collectively impact about 3.5–5.9% of the global population¹. This prevalence highlights the critical importance of investigating rare diseases as a public health priority. Global national initiatives have significantly supported efforts to understand and address these diseases, improve diagnosis rates, and foster research in this domain^{2,3,4,5,6,7,8}. In South Korea, the Korean Undiagnosed Diseases Program (KUDP) initiated the pilot project in 2017⁹, transitioning into its first phase from 2018 to 2020¹⁰. During this phase, it focused on improving diagnostic accuracy and establishing a robust and enduring research infrastructure.

Diagnosing rare diseases in adults presents a greater challenge than in pediatric patients, resulting in many undiagnosed disease programs focusing predominantly on younger populations¹¹. Nonetheless, recent advancements in next-generation sequencing technologies, such as exome sequencing (ES) and genome sequencing (GS), have shown significant potential for diagnosing rare diseases, particularly in adults with neurological conditions^{12,13,14,15,16,17}. The field of genomic medicine is gradually transforming healthcare practices by integrating genomic information into clinical care¹⁸. Despite these advances, the application of ES and GS in diagnosing undiagnosed diseases in adults encounters several obstacles, including the complex task of correlating genotypes with clinical phenotypes, lower diagnostic yields in adults than in children, a greater probability of non-genetic causes, and the absence of effective treatments^19,20,21.

In response to these challenges, the KUDP has recently expanded its mission to enhance efforts toward resolving undiagnosed diseases in adults, marking the beginning of the adult KUDP in 2020. Despite comprehensive evaluations and genetic analyses, the continued modest success rates in diagnosing adults highlight the pressing need for focused research and the development of more effective diagnostic methods²². Herein, we offer critical insights from the adult KUDP to enhance diagnostic processes and push the boundaries of genomic medicine for adults with rare diseases.

Results

Cohort characteristics

A total of 232 probands originating from unrelated families were included, with no significant difference in the sex distribution observed (Table 1). The median age of symptom onset was 30.5 years. They were stratified into two groups based on their clinical profiles: 128 (55.2%) were placed in the “probable genetic origin” group, whereas 104 (44.8%) were categorized into the “uncertain origin” group (Fig. 1). Among the 128 individuals in the probable genetic group, 56 (43.8%) had a family history. Notably, the majority of those without a family history (53/72, 73.6%) presented with neurological disorders, often with atypical early-onset symptoms, which likely raised clinical suspicion of a genetic etiology despite the absence of a known family history. Overall, 170 individuals (73.3%) primarily exhibited neurological disorders, while 62 individuals (26.7%) presented with non-neurological conditions. Within the neurological disorder group, cerebellar ataxia (25.4%) was the most frequently observed phenotype, followed by spastic paraplegia (14.7%) and leukodystrophy (8.6%). In the non-neurological disorders group, musculoskeletal disorders (9.9%) were the most common, followed by cardiovascular diseases (3.4%). Diagnostic genetic testing was extensively utilized, with 226 individuals (97.4%) undergoing ES or GS. A minority of the cohort received either a single gene or targeted panel sequencing (three individuals) or antibody screening (three individuals). The most common sequencing approach was proband-only sequencing, which was used for 172 probands (74.1%) due to the difficulties of obtaining parental samples. However, to enhance the accuracy of variant interpretation, sequencing of additional family members was conducted if possible. As a result, 330 samples from 232 probands and their 98 family members underwent ES or GS, enabling trio sequencing for 34 families (14.7%), duo sequencing for 12 families (5.2%), and quartet sequencing for 8 families (3.4%).

Table 1 Clinical characteristics of adults with undiagnosed rare diseases

Full size table

Implementing genomic medicine in clinical practice for adults with undiagnosed rare diseases — **Fig. 1: Overview of the diagnostic and clinical management workflow for adults with undiagnosed rare disease.**

Non-genetic conditions requiring differential diagnosis with genetic origins

Among 104 individuals in the uncertain-origin group, 12 (11.5%) were ultimately identified as having a non-genetic etiology after receiving appropriate imaging, laboratory studies, immunologic tests, and infection workups as well as genetic testing. Specifically, three individuals with acute muscle weakness and elevated creatine kinase levels were positive for anti-signal recognition particle (SRP) antibodies. ES did not reveal causative variants in them. Their symptoms significantly improved after immunotherapy, which included steroids, intravenous immunoglobulin (IVIG), and rituximab. Another individual exhibited autoantibody against the leucine-rich glioma-inactivated 1 (LGI1) protein, indicating anti-LGI1 encephalitis. Additionally, three other subjects were diagnosed with seronegative autoimmune encephalitis manifesting as acute cognitive decline. Their conditions were characterized by fever, rapid cognitive and motor function decline, and a positive response to immunotherapy, including steroids, IVIG, rituximab, and/or tocilizumab, following the exclusion of infectious and genetic etiologies. Overall, seven individuals had immunological causes.

Another individual presented with subacute dizziness and general weakness and was ultimately diagnosed with fungal meningitis due to Trichosporon asahii²³. This diagnosis was reached after magnetic resonance imaging and cerebrospinal fluid analysis suggested the presence of an atypical pontine infarction and meningitis, with genetic disorders ruled out through ES. Two other individuals with non-genetic causes were concluded to have a rare case of post-pump chorea and a functional movement disorder. Two individuals were diagnosed with Gorham–Stout disease and Parry–Romberg syndrome. Collectively, these cases highlight the intricate challenge of distinguishing non-genetic from genetic causes in undiagnosed adults and underscore the critical role of comprehensive clinical evaluation in identifying non-genetic etiologies that closely overlap with genetic disorders.

Diagnostic rate and strategies to maximize the yield of genetic approach

To enhance the diagnostic rate of genetic tests, we adopted a dynamic approach to data analysis, as illustrated in Fig. 1. We ensured that the databases were routinely updated and that reanalyzes were conducted every six months to incorporate emerging genetic discoveries. This approach led to the identification of causative single nucleotide variants (SNVs) or small indels in 47 individuals (Supplementary Data 2). Notably, 12 (25.6%) received their final diagnosis during the reanalysis phases due to variant reclassification. In addition, we employed multiple bioinformatics tools to detect disease-associated copy-number variations (CNVs) in exome and genome data. CNV detection tools identified disease-associated CNVs in 5 individuals (see Supplementary Table 1 for details). Notably, a deletion affecting the upper regulatory region of the LMNB1 gene was identified (AUDP040) using CONIFER and validated through GS (Fig. 2a). Intriguingly, this was the smallest deletion reported thus far affecting key regulatory domains, and this individual exhibited pronounced leukodystrophy on brain images (Fig. 2b, c)^24,25,26. Additionally, a 1.4 Mb duplication encompassing the PMP22 gene, which is crucial for diagnosing type 1 A Charcot-Marie-Tooth disease, was detected (AUDP109) by CONIFER and confirmed through chromosomal microarray (Fig. 2d). Furthermore, a homozygous deletion spanning exons 1 to 4 of the DRC1 gene was identified (AUDP044) with primary ciliary dyskinesia using HMZDelFinder (Fig. 2e)²⁷. Similarly, a 58.0-kb heterozygous deletion involving the entire SERPINC1 gene was identified (AUDP086) with antithrombin deficiency. In another case, a large inversion including the PAX6 gene was identified using GS in an individual with congenital aniridia (AUDP139).

**Fig. 2: Enhanced detection of clinically significant copy-number variations utilizing bioinformatic tool.**

When autosomal recessive disorders were strongly suspected but only one pathogenic allele was detected, multiplex ligation-dependent probe amplification (MLPA) was employed, as the MPLA proved invaluable for detecting small deletions that have been missed by ES (Supplementary Table 1). This allowed us to detect two small deletions involving SETX exon 26 and PANK2 exon 2. Notably, heterozygous and homozygous alleles of SETX exon 26 deletions were identified in two unrelated individuals (AUDP048 and SUDP059); one (AUDP048) with a heterozygous SETX deletion also possessed a splicing variant (NM_015046.7:c.6106+1 G > A) in the opposite allele, indicating compound heterozygosity²⁸. The detection of the homozygous allele of SETX exon 26 deletion strongly suggested that the SETX exon 26 deletion may be a founder variant in the Korean or East Asian population and could be a frequent cause of ataxia with oculomotor apraxia type 2 (AOA2) in Korea. Similarly, one (SUDP059) with a PANK2 exon 2 deletion had an LP variant (NM_001386393.1:c.803 A > G; p.Asp268Gly) in the opposite allele. Collectively, the above findings highlight the importance of incorporating CNV analysis to optimize diagnostic outcomes.

Considering the high proportions of neurological disorders in our cohort, we extended our evaluation to include short tandem repeat (STR) disorders, which are often missed by standard molecular tests. For this, we screened known STR disorders utilizing exomes (ExpansionHunter)²⁹ and/or conducted RP-PCR or Cas9-mediated nanopore long-read sequencing (LRS), particularly in cases where initial sequencing did not provide definitive results. This extended assessment primarily focused on suspected subgroups with symptoms of cerebellar ataxia or leukodystrophy. As a result, we identified 14 individuals with STR disorders (Supplementary Table 2) within diverse genes^30,31, including NOP56 (n = 5), NOTCH2NLC (n = 4), ATXN3 (n = 2), ATXN7 (n = 1), ATXN8 (n = 1), and ATN1 (n = 1).

Impact of clinical factors on the diagnostic rate

Overall, the disease etiology was identified in 78 probands (33.6%), including genetic origin in 66 (28.4%) and non-genetic origin in 12 (5.2%). All with non-genetic etiologies were found in the uncertain group and included three cases of inflammatory myopathy with anti-SRP antibodies, three of seronegative autoimmune encephalitis, and one each of anti-LGI1 encephalitis, fungal meningitis, post-pump chorea, functional movement disorder, Gorham–Stout disease, and Parry–Romberg syndrome. These conditions highlight the intricate challenge of distinguishing non-genetic from genetic causes in adults with undiagnosed diseases and underscore the critical role of comprehensive clinical evaluation in distinguishing non-genetic etiologies that closely overlap with genetic disorders.

Genetic causes were identified in 66 probands: 51 from the probable genetic origin group and 15 from the uncertain group. In the 15 diagnosed individuals from the uncertain group, differentiating the genetic causes based on initial clinical presentations was particularly challenging, as all were sporadic, with the majority affected by leukodystrophy (n = 6). The final diagnosis was established after identifying causative genetic variants, considering inheritance patterns, conducting familial studies, and thoroughly correlating genotypes with phenotype. Specifically, disease-associated SNVs, CNVs, and STRs (Supplementary Tables 2–4) were identified in 47, 8, and 14 individuals, respectively. These results highlighted the importance of considering diverse genetic variations for diagnostic success.

In sub-group analyses, the diagnostic yield was significantly greater in the probable genetic origin group (39.8%) than in the uncertain origin group (14.4%, P < 0.0001; Fig. 3a). Furthermore, the diagnostic rate was influenced by the presence of family history and age at symptom onset. Specifically, individuals with a family history had a greater diagnostic rate of 42.9% than 23.8% for those without such a history (P = 0.010; Fig. 3b). Despite not reaching statistical significance in our cohort, the age of symptom onset showed a negative relationship with the diagnostic rate (Fig. 3c). Furthermore, our analysis of the relationships between phenotypic categories and diagnostic success revealed no significant correlation across our cohort. However, the diagnostic rate for neurological phenotypes (30.6%) was marginally greater than that for non-neurological phenotypes (22.6%; Fig. 3d). Notably, within the neurological category, individuals who presented with leukodystrophy achieved the highest molecular diagnostic rate at 40.0% (Fig. 3e). This highest yield might be partially attributed to the cases diagnosed by CNV and STR analyses. In contrast, diagnostic rates among non-neurological disorders exhibited a wide variation ranging from 11.1% to 50.0% (Fig. 3f).

**Fig. 3: Diagnostic rates according to clinical factors and phenotype categories.**

The time interval from symptom onset

Next, we analyzed the time interval between the onset of initial symptoms and the achievement of a definitive diagnosis or the last follow-up. Among the 66 individuals who received a genetic diagnosis, the median diagnostic interval was 8.9 years (95% CI: 6.5–12.4 years). In contrast, the undiagnosed individuals had a shorter median follow-up duration of 7.1 years (95% CI: 6.0–8.0 years). However, the difference in intervals between the diagnosed and undiagnosed groups was not statistically significant (Fig. 4a). Additionally, individuals in the probable genetic origin group experienced a longer diagnostic journey compared to those in the uncertain origin group, regardless of whether they received a diagnosis (Fig. 4b). The length of the diagnostic journey did not differ significantly based on the presence or absence of a family history (Fig. 4c). Notably, the higher diagnostic rate observed in individuals with a family history (Fig. 3b) was primarily distinguished during the earlier phase, specifically within the first 10 years of symptom onset (Fig. 4d). This highlights that individuals with a family history are more likely to benefit from early access to genetic testing compared to those without.

**Fig. 4: Impact of clinical factors on the diagnostic journey in adults with rare diseases.**

Additionally, individuals who experienced symptom onset during childhood endured significantly longer diagnostic journeys compared to those whose symptoms began in adulthood, as expected (Fig. 4e). Among the 17 individuals whose symptoms began in childhood and were diagnosed through the adult UDP, we observed that many cases were particularly challenging to diagnose. These included cases requiring the identification of CNVs (n = 3) or two trans alleles for autosomal recessive disorders (n = 6). These complexities may have hindered molecular diagnosis and further prolonged the diagnostic journey. Furthermore, we found that adults with pediatric-onset symptoms initially had a similar or even lower diagnostic rate compared to those whose symptoms began in adulthood. However, this trend reversed in the later phases of the diagnostic journey, particularly around the 30-year mark (Fig. 4f). This suggests that adults might have not fully benefited from recent advancements in molecular technologies during the early stages of their lives, underscoring the critical need for enhanced diagnostic efforts for undiagnosed adults who initially presented with symptoms during childhood.

Personalized management after diagnosis

After receiving molecular diagnoses, 66 individuals were closely monitored for an average duration of 17.2 months, ranging from 0 to 42 months. Each individual, along with their family members, was provided with detailed information about their diseases and related genes supplemented by tailored genetic counseling. For 21 individuals (31.3%), previously unrecognized surveillance measures specific to their genetic conditions were initiated, demonstrating the value of precise genetic diagnosis in enhancing patient care. Additionally, cascade screening was conducted within 10 families (14.9%), resulting in the identification of 12 affected family members and 2 female carriers of X-linked conditions (Supplementary Table 3). Moreover, 10 individuals were identified as suitable candidates for drug repurposing based on their confirmed genetic diagnosis within 9 genes (Table 2). After excluding one who declined treatment, nine consented to receive new treatment options with repurposed drugs. Of these, five individuals experienced benefits from drug repurposing, whereas four did not exhibit definitive responses. Moreover, reproductive counseling was provided to two individuals having autosomal dominant disorders: CACNA1G-related spinocerebellar ataxia and PAX6-related aniridia. They opted for preimplantation genetic testing (PGT) for monogenic/single gene defects (PGT-M) and PGT for structural rearrangements (PGT-SR), which were conducted as preventive measures to ensure the health of their offspring.

Table 2 Cases with drug repurposing following molecular diagnosis

Full size table

For non-genetic causes, personalized treatments were also provided. Three individuals diagnosed with anti-SRP immune-mediated necrotizing myopathy received a treatment regimen of methylprednisolone, IVIG, and rituximab, resulting in significant improvements in muscle weakness and reduced creatine kinase levels following immunotherapy. Similarly, in three individuals diagnosed with autoimmune encephalitis, symptoms such as cognitive decline improved after immunotherapy with rituximab, tocilizumab, and/or IVIG. These outcomes underscore the clinical significance of identifying non-genetic causes in undiagnosed rare diseases, not only for achieving accurate diagnoses but also for tailoring individualized treatment plans.

Discussion

In this study, we enrolled 232 adults with undiagnosed conditions, performed genetic assessments, and analyzed clinical factors affecting diagnostic outcomes and length of diagnostic journeys. Notably, we found a significantly higher diagnostic rate among individuals presumed to have a genetic origin than among those in the uncertain origin group. However, individuals with a probable genetic origin tended to undergo longer diagnostic journeys. Additionally, we successfully identified non-genetic causes in 12 individuals in the uncertain group, emphasizing the critical role of comprehensive clinical evaluations and the difficulties of differential diagnosis in non-genetic causes. Moreover, genomic technologies complementary to short-read sequencing, such as nanopore LRS, have shown extra value in improving the diagnostic process, particularly for the detection of STR disorders associated with cerebellar ataxia and leukodystrophy. Through these integrative approaches, genetic answers were found in 66 (28.4%), comparable to previous reports investigating adults (Supplementary Table 4)^13,15,19. Our findings emphasize the importance of a strategic approach in adults, especially those with known family history and symptoms that commence in childhood, which might potentially shorten the diagnostic odyssey and facilitate personalized care.

This study emphasizes the critical role of continuous reanalysis of genomic data with the latest updates in databases^32,33. During our reanalysis phase, variant reclassification was achieved for 12 cases (25.6%). This highlights the evolving nature of genetic research, where new gene and disease associations are constantly emerging³⁴. Some causative genes and phenotypes were only recognized after thorough literature reviews, stressing the need for ongoing vigilance and expansion of phenotypic associations in the field. Our adult KUDP identified several notable cases of rare diseases, including adult-onset progressive myoclonic epilepsy related to a DHDDS variant³⁵, VEXAS syndrome³⁶, and spinocerebellar ataxia related to CACNA1A (SCA42)³⁷, SPTAN1³⁸, and PRDX3 (SCAR32)³⁹. Additionally, this paper presents a case of leukodystrophy with the smallest noncoding deletion involving the promoter and enhancer region of LMNB1 reported thus far (Fig. 3b)^24,25,26. These discoveries not only enhance the global knowledge of rare diseases but also illuminate the significant contributions of the adult UDP in elucidating the genetic underpinnings of various rare diseases.

Our study highlights the value of advanced genomic technologies, particularly LRS, in diagnosing adults with neurological disorders, especially those involving structural variants and repeat expansions. Using multiple bioinformatics tools^40,41,42, we identified disease-causing CNVs in 5 individuals, representing 7.6% of diagnosed cases (Supplementary Table 1). Additionally, a significant proportion of diagnoses (n = 14, 21.2%) were related to STR disorders, highlighting their major contribution to adult neurological conditions. Notably, Cas9-mediated nanopore LRS was applied to individuals suspected of having repeat expansions in genes such as ATXN7, ATXN8, NOP56, and NOTCH2NLC, demonstrating the unique ability of LRS to analyze genomic regions that are challenging for short-read sequencing. Furthermore, 74.1% of individuals underwent singleton testing, primarily due to the challenges of obtaining parental specimens, which is a common issue among adults. In many cases, parents may be deceased or separated from the individuals, complicating sample collection. LRS offers particular advantages in such scenarios, especially for resolving variant phasing without the need for parental samples, making it a valuable tool for adults. Although not utilized in this study, we have previously demonstrated the feasibility of LRS for variant phasing without the need for parental samples, as exemplified by a case of ataxia-telangiectasia caused by biallelic pathogenic ATM variants⁴³. These findings suggest that future studies utilizing LRS could be expanded to a broader range of STR disorders^44,45,46 and variant phasing challenges, offering more diagnostic opportunities in adult populations.

Early identification of genetic causes is essential, as it facilitates timely medical intervention, accurate risk assessment, and informed family planning for relatives. As such, prioritizing the detection of genetic causes is particularly important for adults with a clear family history (Fig. 3b). In contrast, adults with childhood-onset diseases face a more complex diagnostic journey. Despite a higher diagnostic rate, they experience significantly longer diagnostic timelines, underscoring the need for earlier and more comprehensive genomic analyses (Fig. 4f)^22,33. These prolonged diagnostic journeys, though potentially influenced by recent rapid advancements in sequencing technologies, expose a critical gap in current diagnostic practices. This emphasizes the urgent need for tailored diagnostic strategies to address the specific challenges faced by pediatric-onset individuals when they are evaluated in adult clinics.

We successfully provided personalized management for a subset of adults who received their final diagnoses. Specifically, the process of genetic diagnosis has demonstrated its potential to revolutionize drug repurposing and foster the advancement of personalized medicine, although this potential was realized in a relatively small fraction of cases (14.9%, as shown in Table 2). Nevertheless, the willingness of most individuals to explore drug repurposing was notable, with more than half (5/9) experiencing tangible benefits from the adapted treatments, which highlights the time-saving and cost-effective aspect of drug repurposing in clinical practice⁴⁷. Thus, ongoing and focused medical research into a variety of rare diseases is imperative, and such efforts are essential not only for enhancing our understanding of disease conditions but also for developing targeted therapeutic strategies based on their specific genetic foundations²⁰.

Our cohort was predominantly composed of individuals with neurological disorders (72.8%), a distribution consistent with findings from previous UDP studies in adults (Supplementary Table 4). While the proportion of neurological disorders varies significantly across study cohorts, ranging from 19% to 81%, the composition and diagnostic yield of our study closely align with the findings from the Australian study by Walsh et al.¹⁹. The strength of our study lies in the inclusion of individuals with uncertain etiologies, which led to the identification of non-genetic diagnoses. This highlights the importance of incorporating non-genetic testing and expertise within a UDP program. Despite confirming 12 individuals with non-genetic etiologies through immunological studies or infection workups, the incidence of non-genetic origins in our cohort may be underestimated, as clinicians are more likely to enroll individuals who are expected to benefit from ES or GS. Additionally, potential selection bias may exist due to the use of targeted nanopore LRS for specific genes in a subset of individuals clinically suspected of having certain conditions (cerebellar ataxia and leukodystrophy). Although LRS has shown significant clinical utility for certain phenotypes, individual selection based on clinical suspicion may influence the perceived effectiveness of this technology. Future studies should aim to apply this approach to more diverse cohorts to thoroughly evaluate its potential and limitations. Last, the varied types of previous genetic workups among individuals, ranging from single-gene sequencing to GS or even no genetic tests, complicated direct comparisons of the diagnostic approach within our cohort. Recognizing these limitations, future studies should consider strategies to systematically include and evaluate individuals with both genetic and non-genetic origins of rare diseases to provide a more comprehensive understanding of their etiologies.

Overall, this study emphasizes the essential role of genomic medicine in enhancing clinical practice for adults with undiagnosed rare diseases. Based on our findings, we recommend adopting a structured diagnostic approach for adults, as outlined in Supplementary Figure 1. For cases with a probable genetic or uncertain origin, ES or GS should be utilized as the primary tool. Additionally, LRS can provide additional diagnostic value in ES/GS-negative cases, particularly in detecting CNVs or STRs that may be missed by short-read sequencing methods. Importantly, our findings highlight that genetic evaluations may still be valuable even in cases initially suspected to have a non-genetic cause. The distinction between genetic and non-genetic origins in adults can be challenging, and genetic factors may play an unrecognized role in some cases. Therefore, it is essential that genetic testing remains a consideration, even in individuals where non-genetic causes are suspected. At the same time, specific investigations for non-genetic causes—such as autoimmune screening, infectious disease testing, or imaging studies—should be pursued concurrently when indicated. Integrating advanced sequencing technologies and bioinformatics tools into routine diagnostic workflows is key to overcoming the challenges of diagnosing complex rare diseases. By improving diagnostic accuracy and providing earlier, more precise insights into disease etiology, these approaches can significantly enhance patient care and ultimately lead to better clinical outcomes.

Methods

Study cohort

Between May 2020 and May 2023, we prospectively enrolled a total of 232 probands over 18 years old who were referred to the adult undiagnosed disease clinic in suspicion of undiagnosed rare disease at Seoul National University Hospital and Samsung Medical Center. Referrals were made within the hospital as well as other primary, secondary, and tertiary centers in Korea. Upon referral, individuals’ medical histories, physical examinations, and review of prior diagnostic investigations were comprehensively evaluated in person first by the attending physician and by a multidisciplinary team, which included adult neurologists, medical geneticists, and other specialists relevant to the primary symptoms. Individuals who remained undiagnosed despite relevant diagnostic efforts at other centers, or who presented with symptoms suggestive of rare diseases, were defined as undiagnosed rare diseases and were subsequently enrolled in the study. If the symptoms could be addressed through standard diagnostic processes not requiring a multidisciplinary or genetic approach, as determined by the attending physician, the individual was excluded from the UDP study. Each proband had not undergone prior ES or GS, although more than 70% had received prior targeted genetic tests, and each represented a distinct family. When available, additional family members were included for sequencing to facilitate variant discovery (Supplementary Data 1). All participants were recruited from two tertiary hospitals in Seoul, Republic of Korea: Seoul National University Hospital (SNUH; n = 156) and Samsung Medical Center (SMC; n = 76). Informed consent was obtained from the participants following the Declaration of Helsinki, and the study was approved by the Institutional Review Board of SNUH (IRB no. 2006-083-1132) and SMC (IRB no. SMC-2020-03-025). Collectively, these institutions manage a significant proportion of rare diseases in Korea.

Clinical assessment

Based on the clinical assessments by the attending physician, individuals were classified into two groups based on the presumed origin of their disease: “probable genetic origin” or “uncertain origin (genetic or non-genetic)”. The classification criteria for the “probable genetic origin” group included one or more of the following: (1) disease onset before the age of 18, (2) a family history of the same condition, and (3) exclusion of other organic causes through prior diagnostic efforts. These classifications were subsequently reviewed and discussed by the multidisciplinary team, with consensus reached on the classification of the disease’s origin. Individuals classified as having a “probable genetic origin” underwent appropriate genetic testing, including targeted gene panel testing. For those in the “uncertain origin (genetic or non-genetic)” group, further diagnostic testing was carried out, which included imaging, laboratory studies, immunological tests, and infectious disease workups to rule out non-genetic causes. These investigations were conducted in parallel with genetic testing (Fig. 1). In some cases, new symptoms emerged or additional data became available during the study, leading to a more rapid diagnosis through autoimmune, imaging, or infectious screening. When candidate variants were identified through bioinformatic analysis or by a laboratory physician, additional outpatient clinic visits were arranged to conduct familial sequencing for confirmation of variant origin if possible. In parallel, reverse phenotyping was performed to reassess clinical presentations in the context of the newly identified genetic variants, ensuring that variant pathogenicity was accurately determined.

Exome and GS

A total of 330 samples, comprising 232 probands and 98 family members, were subject to ES or GS. Genomic DNA was extracted from the peripheral blood samples using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). Following extraction, the DNA underwent library preparation using the Agilent SureSelect All Exons V6 hybrid capture kit (Agilent Technologies, Santa Clara, CA, USA) and TruSeq DNA PCR-Free High Throughput Library Prep Kit (Illumina, CA, USA) for ES and GS, respectively. Sequencing was then performed, generating 150-bp paired-end reads on the NovaSeq 6000 platform (Illumina, CA, USA).

Bioinformatics analysis

The sequenced reads were aligned to the human reference genome hg38, which was sourced from the Genomics Public Data on Google Cloud. For robust and reproducible analysis, we utilized the Sarek pipeline implemented in NextFlow^48,49. Specifically, BWA-mem2 was used for read mapping, and DeepVariant and GLnexus were employed for the detection of single-nucleotide variants (SNVs) or small insertions/deletions (indels), and joint genotyping⁵⁰. We applied additional criteria for genotype filtering, including (1) a minimum depth (DP) of at least 8, 2) a Phred-scaled genotype quality (GQ) of 20 or higher, and 3) an allelic balance (AB) between 0.2 and 0.8 for heterozygote calls and above 0.9 for homozygote or hemizygote calls. We employed ANNOVAR for variant annotation against multiple databases, including avsnp153, dbNSFP v4.3a, gnomAD exome v2.1, KOVA2, Online Mendelian Inheritance in Man (OMIM), ClinVar, and SpliceAI^{51,52,53,54,55,56,57}. SNVs were interpreted based on the ACMG-AMP criteria, and individuals with variants classified as pathogenic (P) or likely pathogenic (LP) were considered solved cases (Supplementary Data 2)⁵⁸. Structural variants (SVs) were detected using CoNIFER and HMZDelFinder for exome data and Pariliament2 for genome data, and the functional impact of the identified SVs was predicted using AnnotSV^40,41,42,59. SVs predicted P/LP variants by AnnotSV were manually confirmed according to the American College of Medical Genetics and Genomics (ACMG)/Clinical Genome Resource (ClinGen) criteria (Supplementary Table 1)⁶⁰. We also screened for repeat expansions related to STR disorders using ExpansionHunter v5.0^29,30. Genetic experts interpreted genetic testing results, and clinically significant variants were further validated using orthogonal methods.

Cas9-mediated nanopore LRS

For individuals suspected of having STR disorders, we utilized repeat-primed PCR (RP-PCR) targeting spinocerebellar ataxia (SCA) 10, 12, 31, 36, and 37. Targeted LRS of nanopores, including the ATXN7, ATXN8, NOP56, and NOTCH2NLC genes, was also used as an alternative method to evaluate repeat expansion^30,31. Cas9-mediated target enrichment for the STR regions located in the target genes was carried out as previously described⁶¹ using the gRNAs described in Supplementary Table 5. A total of 5 μg of genomic DNA was processed following the Cas9 sequencing kit protocol (Oxford Nanopore Technologies, UK; SQK-CS9109). The prepared libraries were loaded onto R9.4 flow cells (FLO-MIN107) and sequenced using the MinION platform from Oxford Nanopore Technology. Base calling and FASTQ conversion were performed using MinKNOW v5.3.6. The resulting FASTQ files were aligned to the human reference genome hg38 using minimap2 v2.24. Repeat counts were estimated using the software Straglr v1.4.1⁶².

Statistical analysis

Statistical analyses were conducted using GraphPad Prism v9.5.0. Categorical variables were compared using a two-tailed Fisher’s exact test, and continuous variables were compared using the Mann‒Whitney U-test. The length of the diagnostic period was defined as the period from symptom onset to the date of the last follow-up or the date of diagnosis. A significance level of P < 0.05 was used throughout the analysis.