Current translational perspectives for diagnosis and individualized therapy in rare genetic diseases

Abstract

Rare diseases affect hundreds of millions of individuals worldwide, primarily resulting from genetic factors, with many presenting during childhood. Advances in genome sequencing increasingly enable identification of the responsible gene in more patients; however, only a limited selection of therapies exists to correct or mitigate the underlying genetic defect. This overview delineates the interconnected aspects of diagnosis, evidence generation, treatment, and long-term safety monitoring within an integrated framework. It also posits that the molecular mechanism underlying a disorder, rather than its clinical label, determines both the diagnostic test to be employed and the potential treatment options that may be effective.

Introduction

Rare diseases are distinguished by jurisdiction-specific prevalence thresholds rather than a universal global criterion. The European Union designates a disease as rare if it affects fewer than 5 in 10,000 individuals; simultaneously, the United States’ Orphan Drug Act of 1983 establishes a threshold of fewer than 200,000 individuals nationwide^1,2. The overall impact remains substantial. Approximately 27 to 36 million persons within the European Union are affected by between 6000 and 8000 distinct rare diseases, of which approximately 80% have a genetic origin and about 70% manifest during childhood^1,3. An analysis of the Orphanet database attributed disease-causing germline variations to 71.9% of cataloged conditions, identified pediatric onset exclusively in 69.9%, and estimated the global cumulative point prevalence at 3.5–5.9%⁴. The Rare Diseases International and Lancet Commission estimate the global total at approximately 400 million individuals and underscore significant disparities in diagnosis, treatment, and access.

Diagnostic delay exacerbates the associated burden. A retrospective survey conducted by the European Rare Barometer, involving 6507 individuals across 41 countries, estimated an average delay of 4.7 years from initial consultation to diagnosis, with 56% diagnosed more than six months after the first medical consultation. As a self-selected sample, it is subject to recall and selection biases and does not serve as a population-based measure⁵. Such delays impede preventive surveillance, disease-modifying therapies, reproductive counseling, and eligibility for clinical trials⁶. Patients originating from underrepresented ancestries bear a disproportionate burden of variants of uncertain significance, primarily because reference datasets remain biased toward European populations⁷.

The therapeutic landscape is rapidly transforming. Genome sequencing is increasingly used as a primary diagnostic tool in selected pediatric, neonatal, and undiagnosed cases⁸. Mechanism-targeted therapies now include antisense oligonucleotides, RNA interference, messenger RNA, adeno-associated virus gene replacement, ex vivo lentiviral gene addition, and both ex vivo and in vivo genome editing, with ongoing emergence of new approvals and post-marketing safety signals. In February 2026, the United States Food and Drug Administration issued draft, non-binding guidance outlining considerations for a Plausible Mechanism Framework in developing personalized therapies for genetic conditions with established biological causation⁹. This document is reviewed in relation to its fundamental content within the section on the regulatory environment.

This comprehensive overview considers rare-disease research and care as a unified translational pathway rather than merely a collection of disease-specific initiatives. It encompasses genetic architecture, diagnostic technologies, biomarkers, natural history, shared data infrastructure, preclinical platforms, therapeutic modalities and their safety obligations, trial design—including N-of-1 studies—and equitable access. The principal assertion is that although molecular diagnosis is increasingly attainable, it remains insufficient; transforming a diagnosis into a regulated intervention requires validated gene-disease relationships, context-specific biomarkers, manufacturing feasibility, and comprehensive long-term safety data. The primary constraints are no longer solely technical; they also include evidence standards, regulatory clarity, equitable access, and the recognition that many molecular diagnoses will not lead to available therapies.

Scope and literature selection

Sources were identified purposively rather than systematically. PubMed and Google Scholar were searched from January 2015 to May 2026, with earlier landmark publications retained where they remain definitive, utilizing combinations of terms related to the concept’s rare disease, exome or genome sequencing, diagnostic yield, long-read sequencing, episignature, imprinting, biomarker qualification, gene therapy, genome editing, antisense oligonucleotide, N-of-1, and health equity. Regulatory material was sourced directly from FDA guidance documents, approval letters, prescribing information, and press announcements; EMA public statements; and the Federal Register, supplemented by resources from Orphanet, OMIM, ClinGen, GA4GH, and Genomics England. We prioritized primary studies, formal guidelines, and authoritative registries over secondary commentary, with a preference for evidence with direct diagnostic or regulatory significance. The emphasis reflects the authors’ judgment rather than a predefined protocol. The literature cutoff date was 8 May 2026; regulatory statuses and database identifiers were re-verified against primary sources on 14 August 2026.

Genetic and molecular basis

Architecture and mechanisms

Most cataloged rare genetic disorders are Mendelian, involving autosomal dominant, autosomal recessive, X-linked, and mitochondrial inheritance patterns. Chromosomal and somatic genetic conditions form additional categories. A significant proportion of severe pediatric disorders result from de novo variants that are absent from parental genomes¹⁰. Beyond point mutations, structural variants, repeat expansions, splice-altering and noncoding variants, mosaicism, and digenic or blended phenotypes account for many unresolved cases^11,12. A UK whole-genome study of 802 patients demonstrated that structural and noncoding variants explain a measurable fraction of diagnoses missed by exome or panel testing¹³.

Gene-disease relationships are meticulously curated using evidence frameworks, such as the Clinical Genome Resource, which rates validity on a scale from limited to definitive¹⁴. Variants are systematically classified according to the joint American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology framework, integrating data from population frequency, computational predictions, functional studies, segregation analyses, and allelic information into five hierarchical tiers¹⁵. Both frameworks enhance transparency and facilitate periodic reanalysis.

The mechanism influences both diagnostic and therapeutic strategies. Loss-of-function and haploinsufficiency mechanisms cause numerous deficiencies in enzymes, transcription factors, and structural proteins and are amenable to interventions, such as gene replacement, mRNA supplementation, or paralog upregulation. Addressing gain-of-function and dominant-negative mechanisms involves reducing the pathogenic product through methods including antisense oligonucleotides, RNA interference, or gene editing, either allele-specifically or via non-selective knockdown, with or without subsequent replacement. Misfolding issues may occasionally be remedied with pharmacological chaperones. Enzyme deficiencies that lead to toxic substrate accumulation can be managed through enzyme replacement therapy or substrate reduction strategies. Furthermore, channelopathies often respond to small-molecule modulators.

Two qualifications govern the interpretation of that mapping, which in turn determine the structure of Tables 1 and 2. First, the mechanism does not dictate the diagnostic method. The appropriate initial test depends on the variant class to be identified and the genomic context in which it resides, rather than the pathophysiological category. Exome or genome sequencing detects autosomal dominant loss of function resulting from an intragenic mutation; it does not identify loss of function caused by homozygous deletion of a gene within a segmental duplication. Second, a mechanismally appropriate therapy is not necessarily disease-modifying. Enzyme replacement therapy provides the missing catalytic activity without correcting the underlying genetic defect, and agents acting downstream of a defect, such as MEK inhibitors in neurofibromatosis type 1 or omaveloxolone in Friedreich ataxia, alter consequences rather than the root cause. Consequently, Table 1 reports the mechanism, variant class, inheritance pattern, and diagnostic route as separate axes, whereas Table 2 reports the therapeutic axis separately.

Table 1.

Therapy Mechanisms in Diagnostic Pathway

Molecular mechanism	Representative disorder (OMIM; ORPHA)	Gene or locus	Principal variant class	Inheritance	First-line test	Confirmatory or reflex test
Haploinsufficiency by intragenic variant	Dravet syndrome (OMIM #607208; ORPHA:33069)	SCN1A	Point and small indel; intragenic deletion	Autosomal dominant, mostly de novo	Epilepsy gene panel or trio exome sequencing	Dosage assay for intragenic deletion; RNA analysis for splice variants
Haploinsufficiency by intragenic variant	Neurofibromatosis type 1 (OMIM #162200; ORPHA:636)	NF1	Point, splice, and intragenic copy-number change	Autosomal dominant	Exome or genome sequencing with copy-number calling	MLPA for whole-gene and multiexon deletion; RNA analysis for deep intronic variants
Loss of function by homozygous deletion within a segmental duplication	5q spinal muscular atrophy (ORPHA:70; type I OMIM #253300)	SMN1 with SMN2 copy-number modifier	Homozygous exon 7 deletion or gene conversion	Autosomal recessive	SMN1 exon 7 dosage by MLPA or quantitative PCR, with SMN2 copy number	SMN1 sequencing in the approximately 5% who are compound heterozygous. Exome and standard short-read genome pipelines do not reliably resolve SMN1 without a paralog-aware caller
Lysosomal enzyme deficiency	Gaucher disease type 1 (OMIM #230800; ORPHA:77259); Pompe disease (OMIM #232300; ORPHA:365); mucopolysaccharidosis type I (OMIM #607014; ORPHA:579)	GBA1, GAA, IDUA	Missense, null, and recombinant alleles	Autosomal recessive	Leukocyte or dried-blood-spot enzyme activity; urine glycosaminoglycans for mucopolysaccharidosis	Gene sequencing with a pseudogene-aware method for GBA1; genotype for prognosis and therapy selection
Loss of enzyme function with substrate accumulation (aminoacidopathy)	Phenylketonuria (OMIM #261600; ORPHA:716)	PAH	Point and splice variants	Autosomal recessive	Newborn screening phenylalanine measurement	PAH sequencing; tetrahydrobiopterin responsiveness testing
Loss of dystrophin function	Duchenne muscular dystrophy (OMIM #310200; ORPHA:98896)	DMD	Exonic deletion or duplication in about 70%; small variants; deep intronic splice variants	X-linked; most heterozygous females are asymptomatic carriers, while a minority develop cardiomyopathy or muscle symptoms	MLPA or array comparative genomic hybridization for exonic dosage	Gene sequencing for small variants; RNA analysis or muscle dystrophin studies for splice-disrupting variants
Toxic gain of function with protein aggregation	Hereditary transthyretin amyloidosis (OMIM #105210; ORPHA:85447); SOD1-related amyotrophic lateral sclerosis (OMIM #105400; ORPHA:803)	TTR, SOD1	Missense variants	Autosomal dominant	Targeted panel or single-gene sequencing	Tissue biopsy with amyloid typing; cardiac imaging and scintigraphy for the cardiac phenotype
Repeat expansion causing transcriptional silencing	Friedreich ataxia (OMIM #229300; ORPHA:95); fragile X syndrome (OMIM #300624; ORPHA:908)	FXN, FMR1	Intronic GAA expansion; 5’ UTR CGG expansion with methylation	Autosomal recessive (FXN); X-linked (FMR1)	Repeat-primed PCR; methylation analysis for FMR1	Long-read sequencing for accurate sizing and interruption mapping; FXN sequencing in compound heterozygotes
Repeat expansion causing RNA toxicity	Myotonic dystrophy type 1 (OMIM #160900; ORPHA:273)	DMPK	3’ UTR CTG expansion	Autosomal dominant with anticipation	Repeat-primed PCR	Southern blot or long-read sizing for large alleles and somatic mosaicism
Repeat expansion producing a toxic protein	Huntington disease (OMIM #143100; ORPHA:399)	HTT	Coding CAG expansion	Autosomal dominant	Repeat-primed PCR with fragment sizing	Confirmatory sizing; predictive testing only under a counseling protocol
Recurrent contiguous-gene deletion	22q11.2 deletion syndrome (OMIM #188400; ORPHA:567); Williams-Beuren syndrome (OMIM #194050; ORPHA:904); proximal 16p11.2 deletion syndrome (OMIM #611913; ORPHA:261197)	22q11.2, 7q11.23, 16p11.2	Recurrent microdeletion mediated by segmental duplications	De novo in most probands; autosomal dominant when inherited	Exome or genome sequencing with validated copy-number calling	Chromosomal microarray where such sequencing is unavailable, or where a specific recurrent syndrome is strongly suspected; optical genome mapping for unresolved balanced events
Mitochondrial dysfunction, nuclear-encoded	TK2 deficiency (OMIM #609560; ORPHA:254875); Barth syndrome (OMIM #302060; ORPHA:111)	TK2, TAFAZZIN	Point and small indel variants	Autosomal recessive (TK2); X-linked (Barth)	Nuclear gene panel or exome or genome sequencing	Enzymology; mtDNA copy-number quantification in muscle; monolysocardiolipin for Barth syndrome
Mitochondrial dysfunction, mtDNA-encoded	MELAS (OMIM #540000; ORPHA:550). Leigh syndrome (OMIM #256000; ORPHA:506) is genetically heterogeneous and may be nuclear or mtDNA-encoded	mtDNA, commonly m.3243 A > G (MT-TL1)	mtDNA point variant or single large deletion	Maternal, with heteroplasmic segregation	mtDNA sequencing from urinary epithelium or muscle rather than blood alone	Heteroplasmy quantification in a clinically affected tissue; respiratory-chain enzymology
Loss of parent-of-origin specific expression (imprinting)	Prader-Willi syndrome (OMIM #176270; ORPHA:739); Angelman syndrome (OMIM #105830; ORPHA:72); Beckwith-Wiedemann syndrome (OMIM #130650; ORPHA:116); Silver-Russell syndrome (OMIM #180860; ORPHA:813)	15q11-q13, 11p15	Deletion, uniparental disomy, imprinting-center defect, or single-gene variant (UBE3A, CDKN1C)	Not Mendelian in most cases; recurrence risk depends on the mechanism	Methylation-specific MLPA, which reports methylation and dosage together and can therefore identify a deletion	SNP array or microsatellite analysis to distinguish uniparental disomy from an imprinting-center epimutation; UBE3A or CDKN1C sequencing
Disruption of the epigenetic machinery	Kabuki syndrome type 1 (OMIM #147920; ORPHA:2322)	KMT2D, KDM6A	Point, nonsense, and frameshift variants	Autosomal dominant, mostly de novo	Exome or genome sequencing	Episignature classifier to resolve variants of uncertain significance
Structural or quantitative globin defect	Sickle cell disease (OMIM #603903; ORPHA:232); transfusion-dependent beta-thalassemia (OMIM #613985; ORPHA:848)	HBB	NM_000518.5 (HBB): c.20 A > T (p.Glu7Val); point and deletional beta-thalassemia alleles	Autosomal recessive	Newborn screening, electrophoresis or HPLC	HBB sequencing; deletion analysis of the alpha- and beta-globin gene clusters

Note. Each disorder appears once, under the mechanism that defines its pathogenesis. HPLC high-performance liquid chromatography, MELAS mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes, MLPA multiplex ligation-dependent probe amplification, mtDNA mitochondrial DNA, TK2 thymidine kinase 2, UTR untranslated region.

Table 2.

Mechanism-directed therapy for the disorders in Table 1

Disorder	Therapeutic logic	Representative agent (FDA status)	Principal limitation
Dravet syndrome	Increase productive SCN1A transcript or modulate downstream excitability	None approved; splice-modulating ASO in clinical development	No agent yet corrects the haploinsufficiency; current care is antiseizure medication
Neurofibromatosis type 1	Inhibit RAS-MAPK signaling downstream of neurofibromin loss	Selumetinib (approved for adults, 19 November 2025)74,75; mirdametinib (11 February 2025)76	Acts downstream of the gene defect; treats plexiform neurofibroma only; cardiac, ocular, and cutaneous toxicity
5q spinal muscular atrophy	Replace SMN1 or increase SMN2 exon 7 inclusion	Onasemnogene abeparvovec-xioi and intrathecal onasemnogene abeparvovec-brve (Itvisma, 24 November 2025, patients aged 2 years and older)77,78; nusinersen; risdiplam42	Boxed Warning for serious liver injury; irreversible motor-neuron loss limits late benefit; ASO requires repeated intrathecal dosing
Lysosomal enzyme deficiencies	Supply the missing enzyme, reduce substrate load, or replace hematopoietic cells with corrected cells	Imiglucerase; alglucosidase alfa; laronidase; miglustat; atidarsagene autotemcel for metachromatic leukodystrophy	Does not correct the gene defect; poor CNS penetration; lifelong infusion; anti-drug antibodies; conditioning toxicity for ex vivo approaches
Phenylketonuria	Restrict substrate or augment residual enzyme activity	Dietary therapy; sapropterin in responsive genotypes	Lifelong adherence; responsiveness is genotype-dependent
Duchenne muscular dystrophy	Restore an in-frame dystrophin transcript or deliver microdystrophin	Exon-skipping ASOs; delandistrogene moxeparvovec, restricted to ambulatory patients aged 4 years and older55,56	Boxed Warning for acute liver injury; indication narrowed November 2025; exon-skipping benefit is mutation-specific and modest
Hereditary transthyretin amyloidosis	Reduce or stabilize the amyloidogenic protein	Patisiran, vutrisiran, inotersen, eplontersen; tafamidis and acoramidis for the cardiac phenotype	Does not reverse established amyloid deposition; inotersen carries a Boxed Warning and a REMS
SOD1-related amyotrophic lateral sclerosis	Knock down mutant SOD1 transcript	Tofersen (accelerated approval April 2023 based on plasma neurofilament light chain)79,80	Confirmatory clinical-benefit verification outstanding; benefit limited by motor-neuron loss already sustained
Friedreich ataxia	Activate Nrf2 to mitigate downstream mitochondrial and redox consequences	Omaveloxolone (February 2023, patients aged 16 years and older)81	No effect on the causal repeat expansion; transaminase elevation; benefit measured on a rating scale over 48 weeks
Fragile X syndrome	Reverse transcriptional silencing of FMR1 and restore FMRP expression	None approved; epigenetic editing and reactivation strategies remain preclinical	No agent reverses established promoter methylation, and the developmental window may limit attainable benefit
Myotonic dystrophy type 1	Degrade or sequester the expanded CUG-repeat transcript and release sequestered splicing factors	None approved; repeat-targeted antisense oligonucleotide and small-molecule approaches investigational	Delivery to skeletal muscle and the multisystem phenotype remain unsolved; somatic repeat instability complicates response prediction
Huntington disease	Lower total or allele-selective mutant HTT transcript and protein	None approved; HTT-lowering antisense oligonucleotide, RNA interference, and base-editing programs in clinical development	Dose-dependent adverse effects have been reported in trials; allele-selective targeting is technically demanding and predictive testing carries counseling obligations
Recurrent contiguous-gene deletion syndromes	No mechanism-directed therapy for a multigene deletion	Multidisciplinary surveillance and symptomatic care	Gene-specific replacement is investigational and addresses only one gene within the interval
Mitochondrial disorders	Supply a deficient nucleoside, stabilize the inner membrane, or support residual respiratory function	Doxecitine and doxribtimine (Kygevvi, 3 November 2025), for patients with symptom onset at 12 years of age or younger82; elamipretide (Forzinity, accelerated approval 19 September 2025), for patients weighing at least 30 kg83	Confirmatory-trial obligations; no approved therapy for mtDNA-encoded disease; heteroplasmy complicates response prediction
Imprinting disorders and Mendelian disorders of the epigenetic machinery	Restore expression from the silenced allele or correct the chromatin defect	None approved; ASO and methylation-modulating approaches investigational	Symptomatic and surveillance-based care remains the standard
Sickle cell disease and transfusion-dependent beta-thalassemia	Induce fetal hemoglobin or add a functional beta-globin gene	Exagamglogene autotemcel50,51; lovotibeglogene autotemcel	Myeloablative conditioning; infertility; Boxed Warning for hematologic malignancy on lovotibeglogene autotemcel; 15-year follow-up obligation

Note. ASO antisense oligonucleotide, CNS central nervous system, REMS Risk Evaluation and Mitigation Strategy, TK2 thymidine kinase 2.

Status is FDA status as of 8 May 2026, re-verified 14 August 2026, and is jurisdiction-dependent; multinational authorizations are listed in Supplementary Table S2

Genotype does not entirely predict phenotype, as factors such as expressivity, penetrance, modifier genes, environment, age, sex, and ancestry-specific data all complicate counseling and trial design¹⁶. Patients with identical variants can differ in onset, severity, and organ involvement; therefore, longitudinal phenotyping within structured vocabularies is important. The Human Phenotype Ontology, comprising over 18,000 hierarchically organized terms, is suitable for computational analysis^16,17. (Tables 1 and 2)

Table 1 provides information on the molecular mechanisms and diagnostic routes in representative rare genetic diseases. Disease definitions adhere to OMIM and Orphanet, verified as of 14 August 2026^18,19. The mechanism does not determine the diagnostic pathway; these are reported as separate axes. Table 2 presents therapies independently.

Diagnostic pipeline: from clinical suspicion to molecular confirmation

Clinical recognition and deep phenotyping

Diagnosis begins with systematic recognition. The phenotypic profile includes family history, multisystem examination, growth and developmental patterns, age at onset, and biochemical, hematologic, electrophysiologic, and imaging findings. Documenting this profile using standardized Human Phenotype Ontology terminology enables computational variant prioritization and federated patient matching across institutions¹⁷.

Established genetic testing

The diagnostic approach has transitioned from chromosomal microarray and targeted panels to whole-exome sequencing (WES) and, increasingly, whole-genome sequencing (WGS). The American College of Medical Genetics and Genomics (ACMG) evidence-based guidelines strongly advocate for exome or genome sequencing as either a first- or second-tier diagnostic test in pediatric patients presenting with congenital anomalies, developmental delays, or intellectual disabilities, citing higher diagnostic yield, favorable clinical utility, and acceptable cost-effectiveness when employed early²⁰. This overview consistently endorses that recommendation: for these indications, exome or genome sequencing with validated copy-number calling is the preferred initial or secondary testing modality. Chromosomal microarray retains an established role, particularly when sequencing is unavailable, turnaround time or cost is a concern, or clinical features strongly suggest a specific recurrent microdeletion syndrome^21,22. Outside these scenarios, the choice of diagnostic modality depends on phenotype, urgency, suspected mechanism, previous results, and the capacity of local laboratories and genetic counseling services. Two classes of mechanisms are entirely outside the default sequencing-first approach. Copy-number variations within segmental duplications, such as 5q spinal muscular atrophy, require a locus-specific dosage assay because paralog homology impairs standard short-read alignment. Imprinting defects require methylation analysis since the causal lesion may involve no change in DNA sequence.

Trio sequencing improves de novo variant detection, supports phase determination for compound heterozygosity, and refines variant filtering. Rapid genome sequencing in neonatal and pediatric intensive care units provides actionable results within a few days and, as shown by randomized and observational studies, influences clinical management^23,24.

Diagnostic yield varies significantly by phenotype, age, prior testing, platform, and pipeline. A comprehensive meta-analysis of 71 studies involving over 13,000 individuals reported unweighted mean yields of approximately 45% for initial genome sequencing and approximately 33% in cohorts with prior testing⁸. In the multicenter GREGoR cohort, genomic evaluation diagnosed 218 of 744 families (29.3%), and 61 families (8.2% of the cohort) had variants that specifically required genome sequencing, including structural, deep intronic, and tandem repeat variants²⁵. The United Kingdom 100,000 Genomes pilot project reported an overall diagnostic yield of approximately 25%, increasing to 40–55% in cases involving intellectual disability, hearing, and vision disorders²⁶. Table 3 summarizes comparative diagnostic yields by technology, and Supplementary Table S4 provides detailed information.

Table 3.

Diagnostic technologies in rare-disease workup

Technology	Principal use	Key limitation	Indicative yield
First-line testing
Targeted gene panel	Well-defined syndromic phenotype	Restricted to listed genes	10 to 40%6,8
Targeted dosage assay (MLPA, quantitative PCR)	Loci that short-read sequencing cannot resolve, notably within segmental duplications	Detects only the change it is designed to measure	Test of record for 5q SMA and for exonic dosage in DMD and NF1
Chromosomal microarray	Copy-number variants and large unbalanced rearrangements	Misses balanced events, sequence variants, repeats, methylation defects	15 to 20% in unexplained developmental delay21,22
Whole-exome sequencing (trio)	Broad coding-genome interrogation	Misses noncoding variants, most structural variants, repeats, and paralogous loci	25 to 40%6
Whole-genome sequencing	First- or second-tier test in pediatric congenital anomalies, developmental delay, or intellectual disability20	Noncoding interpretation difficult; paralogous loci need dedicated callers	Approximately 45% first-line and 33% incremental after a negative exome8,25,26
Methylation-specific MLPA and targeted imprinting assays	Imprinting disorders	Cannot separate uniparental disomy from an imprinting-center epimutation; misses low-level mosaicism	Standard of care for Prader-Willi, Angelman, Beckwith-Wiedemann, and Silver-Russell syndromes
Problem-solving after a negative first-line result
Long-read sequencing	Structural variants, repeat sizing, direct methylation reading, phasing	Cost; maturing clinical pipelines	7 to 17% incremental84,85
Optical genome mapping	Megabase-scale and balanced rearrangements	Cannot detect sequence variants or small indels	Adjunct; full concordance with karyotyping, FISH, and microarray in an 85-patient validation86
RNA sequencing	Splicing validation and allele-specific expression	Requires an informative tissue	5 to 35% incremental over exome6
Episignature classifiers	Mendelian disorders of the epigenetic machinery; variant reclassification	Returns inconclusive results outside the reference set	Validated signatures for more than 100 conditions; 18.7% in screening and 32.4% in targeted analysis27,28
Untargeted metabolomics and proteomics	Pathway-level readout supporting pathogenicity	Sparse reference ranges; limited cross-laboratory standardization	Contributed to variant interpretation in 74 of 170 patients (43.5%) and confirmed diagnosis in 21 (12.3%)87–89
Functional assays (cell, organoid, model organism)	Causality and PS3 or BS3 evidence	Time- and resource-intensive	Selective use for variants of uncertain significance14,30

Note. FISH fluorescence in situ hybridization, MLPA multiplex ligation-dependent probe amplification, SMA spinal muscular atrophy.

Yields vary by phenotype, age, prior testing, family structure, and analytical pipeline; full comparative data are given in Supplementary Table S4

Emerging and investigational omics approaches

Long-read sequencing, RNA sequencing, methylation profiling, and metabolomic and proteomic analyses elucidate a considerable proportion of cases that yield negative results with short-read methodologies. Long-read platforms can identify structural-variant breakpoints at single-nucleotide resolution, measure pathogenic repeat expansions at loci, such as FXN, FMR1, RFC1, FGF14, DMPK, and HTT, detect 5-methylcytosine without bisulfite conversion, resolve complex rearrangements, and phase haplotypes without parental samples. Reported incremental diagnostic yields range from 7 to 17% across various cohorts. Optical genome mapping functions as an adjunct technique for resolving structural and balanced rearrangements that remain undetermined. RNA sequencing verifies splicing effects and facilitates the reclassification of variants of uncertain significance, with diagnostic yield contingent upon tissue type and target gene.

Methylation-based testing requires a more precise statement than is usually provided, as the disorders it addresses do not share a common mechanism. Imprinting disorders have a common endpoint: the loss or gain of parent-of-origin-specific methylation at an imprinted locus. This endpoint can be achieved through at least four distinct pathways: deletion, uniparental disomy, an imprinting-center defect, and a pathogenic variant in an imprinted gene, such as UBE3A in Angelman syndrome or CDKN1C in Beckwith-Wiedemann syndrome. Beckwith-Wiedemann syndrome exemplifies this spectrum, resulting from either a loss of methylation at the centromeric imprinting center, gain of methylation at the telomeric center, commonly mosaic paternal uniparental disomy, or a CDKN1C variant. These pathways differ in recurrence risks, implications for tumor surveillance, and counseling considerations. Consequently, an abnormal methylation result indicates the initiation of mechanism identification rather than a definitive diagnosis. Methylation-specific MLPA assesses methylation status and gene dosage simultaneously and can detect deletions; however, it cannot distinguish uniparental disomy from an imprinting-center epimutation on its own. SNP arrays or microsatellite analysis resolve this distinction, while sequencing identifies the specific single-gene subgroup.

Mendelian disorders of the epigenetic machinery constitute a distinct category. These disorders are conventional single-gene anomalies affecting chromatin regulators that manifest a reproducible genome-wide methylation signature as a downstream effect. Episignature classifiers identify this signature, rather than an imprinting defect. The EpiSign v5 classifier possesses validated signatures for over 100 conditions. A study evaluating clinical impact reported a positivity rate of 18.7% in comprehensive screening and 32.4% in targeted analysis of cases with variants of uncertain significance^27,28.

A frequently overlooked limitation across these methodologies is the general inability to detect rare exceptions during routine application of the evaluated assays. Such exceptions include low-level mosaicism, small or atypical deletions within imprinting centers, balanced rearrangements disrupting imprinted domains, and tissue-specific epigenetic modifications, all of which typically fall below the detection thresholds of standard procedures. Furthermore, episignature classifiers tend to produce inconclusive results rather than definitive negatives for conditions and populations absent from their reference cohorts. While a negative result effectively rules out the common pathogenic mechanisms associated with the disorder under investigation, it does not entirely exclude the presence of the disorder itself. Additionally, when clinical suspicion remains high, extended or repeat testing on alternative tissue samples is justified. Untargeted metabolomics and proteomics provide pathway-level information that increasingly helps interpret variants in unresolved syndromic cases.

Multi-omics, AI-assisted interpretation, reanalysis, and functional validation

Integrating genomic, transcriptomic, metabolomic, proteomic, epigenomic, and clinical data provides more compelling evidence of pathogenicity than any individual modality. Federated case matching via the Matchmaker Exchange has facilitated the discovery of novel gene-disease associations across cohorts with overlapping phenotypes²⁹. Artificial intelligence applications in phenotype encoding, variant prioritization, and case identification within electronic health records offer tangible benefits when properly validated; such validation necessitates external validation across diverse ancestries, calibrated measures of uncertainty, model auditability, clinician oversight, and prospective evidence of clinical utility. Periodic reanalysis, driven by updated curation, revised ACMG criteria, expanded population references, and enhanced predictors for noncoding regions and splicing, contributes to diagnoses in 5–15% of unresolved academic cases every 12–24 months^6,25. Functional validation conducted in disease-relevant cells, induced pluripotent stem cells (iPSCs), or organoids, with controls, replication, and blinded analysis, bolsters PS3 or BS3 evidence and often proves decisive for reclassification³⁰. Genomic testing also creates obligations around incidental and secondary findings, which affect relatives as well as probands; consequently, consent processes should include reanalysis and data retention policies. (Table 3 and Fig. 1)

Fig. 1.

Diagnostic and translational workflow for suspected rare genetic disease. Structured phenotyping guides first-line genomic testing and standardized variant interpretation. Unresolved cases proceed to multi-omics reanalysis and functional validation. A confirmed molecular diagnosis enables mechanism-based therapy, registry enrollment, and trial matching, with reanalysis at 12 to 24 months. HPO Human Phenotype Ontology, iPSC induced pluripotent stem cell, LP likely pathogenic, VUS variant of uncertain significance, WES whole-exome sequencing, WGS whole-genome sequencing.

Biomarkers, natural history, and shared data infrastructure

Biomarkers and the BEST framework

The FDA-NIH BEST resource delineates biomarker categories pertinent to rare diseases, including diagnostic, prognostic, predictive, pharmacodynamic and response, monitoring, susceptibility and risk, and safety biomarkers. Additionally, it encompasses surrogate endpoints, digital and imaging biomarkers, and patient-centered outcome assessments³¹. Three principles govern their use. Regulatory status is tied to a specific context of use defined by population, condition, and intended application; it is not attributed to the biomarker in the abstract. Consequently, a biomarker qualified in one context does not automatically qualify in another. Validation occurs across three distinct levels: analytical validity, which refers to measurement accuracy and reproducibility; clinical validity, which pertains to the association with disease state or progression; and clinical utility, which indicates demonstrated improvements in decision-making or health outcomes. Qualification via the FDA Drug Development Tool program or equivalent EMA mechanisms allows reuse across different programs for the same context of use without requiring requalification^32,33. Most proposed biomarkers for rare diseases possess partial evidence, generally demonstrating analytical validity and a plausible mechanistic link. Such biomarkers should consequently be described within those parameters. Supplementary Table S3 comprehensively outlines the categories, representative sources, and validation requirements.

Natural history and patient-centered outcomes

Many rare diseases lack sufficient natural history data, complicating endpoint selection and biomarker validation. The Food and Drug Administration (FDA) guidance emphasizes the importance of natural history data for trial design and interpretation in small populations and advocates for prospective data collection whenever feasible^2,34. It is crucial to explicitly characterize longitudinal stability and the influence of factors, such as age, sex, ancestry, and disease stage. Furthermore, a surrogate endpoint must establish a defensible linkage to clinical benefit³². Patient-centered outcome measures should accurately reflect what patients and their families find meaningful, including mobility, communication, seizure burden, fatigue, feeding, sleep, pain, cognition, independence, and caregiver burden. When validated instruments are unavailable, sponsors and patient organizations often develop new tools, frequently through FDA qualification programs.

Patient registries and federated data ecosystems

Small, dispersed, heterogeneous populations require registries as essential infrastructure. Effective registries use precise case definitions with molecular confirmation, collect longitudinal data at meaningful intervals, develop core outcome sets with patient input, link to biospecimens, and adhere to the FAIR principles of findability, accessibility, interoperability, and reusability³⁵. Orphanet maintains a nomenclature for rare diseases and ORPHAcodes¹⁹; the European Platform on Rare Disease Registration addresses fragmentation among national registries³⁶; the Rare Disease Cures Accelerator-Data and Analytics Platform provides a precompetitive environment for standardized data sharing, disease modeling, and endpoint development³⁷; the Matchmaker Exchange facilitates federated gene and phenotype matching²⁹; and the Global Alliance for Genomics and Health establishes interoperability standards, including phenopackets³⁸. Supplementary Table S1 catalogs these resources.

The federation model, where data are queried in situ rather than centralized, has become the predominant approach for cross-border analysis. It maintains local custodianship while enabling cohort-level research. Governance must include consent for recontact and reanalysis, compliance with the European Union General Data Protection Regulation during cross-border data transfer, implementation of dynamic consent frameworks, adoption of privacy-preserving computational methods, protocols for transitioning from pediatric to adult care, terms governing commercial access, and equitable benefit-sharing with patient communities. The underrepresentation of low- and middle-income countries in existing registries reflects a structural disparity that requires continued investment³⁹. (Fig. 2)

Fig. 2.

The rare-disease data ecosystem. Patients, families, and their genotypic and phenotypic data sit at the center of this ecosystem, surrounded by clinical centers, diagnostic laboratories, patient organizations, research networks, variant knowledge bases, registries, matchmaker networks, biobanks, sponsors, regulators, and payers. Implementing shared standards enables federated, privacy-preserving analysis across international boundaries. ERN denotes European Reference Network, HTA signifies health technology assessment, UDN refers to the Undiagnosed Diseases Network.

Therapeutic modalities and safety obligations

Conventional, small-molecule, and nucleic-acid therapies

Established therapeutic approaches include enzyme replacement therapy, substrate reduction therapy, pharmacological chaperones, cofactor or vitamin therapy, small-molecule pathway modulators, immunomodulation, dietary therapy for inborn errors of metabolism, and repurposing existing drugs. Enzyme replacement therapy has considerably transformed the management of multiple lysosomal storage disorders, subject to two limitations outlined in Table 2: large molecules are incapable of crossing the blood-brain barrier, thereby necessitating alternative strategies for neuronopathic forms, and enzyme replacement provides activity without correcting the underlying genetic defect, thus neither reversing established fibrosis nor eliminating the need for lifelong treatment. Subtypes responsive to cofactors may improve significantly with cost-effective supplements, provided they are correctly identified.

Antisense oligonucleotides modulate pre-mRNA splicing, mRNA stability, or translation and are suitable for rare diseases because they can target specific defects, including patient-specific variants. Nusinersen, an intrathecal oligonucleotide promoting SMN2 exon 7 inclusion, demonstrated improved outcomes in spinal muscular atrophy within the ENDEAR and CHERISH trials^40,41, while risdiplam extended the same splicing mechanism to an oral formulation⁴². Milasen, a splice-switching oligonucleotide developed for a single patient with CLN7 neuronal ceroid lipofuscinosis (OMIM #610951; ORPHA:228366), is discussed below in the N-of-1 literature⁴³. RNA interference offers durable suppression of pathogenic transcripts in hereditary transthyretin amyloidosis, primary hyperoxaluria type 1, and acute hepatic porphyria^44–48. mRNA replacement therapy remains investigational for inborn errors of metabolism, where transient expression may be sufficient.

Gene replacement, gene editing, and cell therapy

Adeno-associated virus (AAV) vectors are employed to deliver functional genes to specific tissues in monogenic loss-of-function disorders. Voretigene neparvovec, indicated for biallelic RPE65 retinal dystrophy, demonstrated improvements in functional vision during phase 3 clinical trials and holds the distinction of being the first FDA-approved AAV gene therapy for an inherited retinal disease. Onasemnogene abeparvovec, a single-dose intravenous therapy utilizing AAV9 that replaces the SMN1 gene, has been shown to enhance survival rates and motor milestones. It is now incorporated into newborn screening programs across multiple regions. An intrathecal formulation for patients aged two years and above received authorization on 24 November 2025, supported by outcomes from the phase III STEER and phase IIIb STRENGTH studies. This formulation includes a Boxed Warning for severe liver injury, requires peri-infusion corticosteroids, and mandates hepatic monitoring. Ex vivo lentiviral gene addition therapies for conditions, such as metachromatic leukodystrophy, transfusion-dependent beta-thalassemia, sickle cell disease, and cerebral adrenoleukodystrophy involve autologous hematopoietic stem-cell modification with myeloablative conditioning, complex manufacturing processes, and intensive post-infusion surveillance.

Genome editing has advanced from preclinical validation to commercial product approval. Exagamglogene autotemcel is a non-viral, ex vivo CRISPR-Cas9-edited autologous CD34+ therapy that targets the GATA1 binding site within the +58 erythroid-specific BCL11A enhancer, thereby de-repressing gamma-globin and restoring fetal hemoglobin levels⁴⁹. The CLIMB SCD-121 trial demonstrated a sustained absence of severe vaso-occlusive crises for at least 12 months in 29 of 30 evaluable patients⁵⁰, whereas CLIMB THAL-111 reported transfusion independence in 32 of 35 evaluable patients⁵¹. The United States Food and Drug Administration (FDA), the Medicines and Healthcare products Regulatory Agency (MHRA), and the European Medicines Agency (EMA) authorized the product for both indications between November 2023 and February 2024, with a stipulation for a 15-year follow-up.

As of 8 May 2026, FDA-approved cell and gene therapies for rare genetic diseases encompass inherited retinal dystrophy, spinal muscular atrophy, hemophilia B, and hemophilia A until the withdrawal described below. These therapies also include metachromatic leukodystrophy, recessive dystrophic epidermolysis bullosa, Duchenne muscular dystrophy, sickle cell disease, transfusion-dependent beta-thalassemia, and AADC deficiency, the last through the first intraputaminal gene therapy⁵². In April 2026, Otarmeni (lunsotogene parvec-cwha) received accelerated approval as the inaugural gene therapy for genetic hearing loss, addressing severe-to-profound and profound sensorineural hearing loss at any frequency greater than 90 dB associated with molecularly confirmed biallelic OTOF variants, in patients with preserved outer hair cell function and no prior cochlear implant in the same ear. As with other accelerated approvals, ongoing authorization may depend on confirmatory evidence⁵³. Approval does not guarantee availability: BioMarin withdrew valoctocogene roxaparvovec worldwide on 23 February 2026 for commercial reasons, which the sponsor stated were unrelated to safety or efficacy⁵⁴. Supplementary Table S2 provides details on products, indications, and jurisdictional status.

Treatment selection and therapeutic futility

Treatment selection depends on factors, such as mechanism, tissue accessibility, reversibility of damage, therapeutic window, and measurable outcomes. A therapy aimed at restoring a missing protein is most effective when administered before irreversible damage occurs. Liver-directed gene therapy will not address neurological manifestations if the agent cannot traverse the blood-brain barrier. Gene replacement therapy is suitable for stable loss of function, whereas toxic gain of function is more effectively managed through allele-specific silencing or editing. Clinicians must explicitly recognize futility. Once motor neurons are lost, photoreceptors have degenerated, fibrosis has replaced functional tissue, or developmental milestones have been irreversibly missed, a mechanism-appropriate therapy may still fail to produce outcomes that are meaningful to patients. Factors, such as disease stage, tissue reserve biomarkers, and transparent discussions of attainable and unattainable benefits are vital to shared decision-making.

Safety obligations and durability

Cell and gene therapies possess distinct safety obligations, categorized by modality as outlined in Table 4. AAV products have been linked to hepatotoxicity, thrombotic microangiopathy, complement activation, troponin elevation, and infrequent fatalities. These risks depend on vector dose, immune status, age, baseline hepatic condition, and corticosteroid management. In the aftermath of two fatal cases of acute liver failure in non-ambulatory boys with Duchenne muscular dystrophy, the Food and Drug Administration (FDA) issued a Boxed Warning in November 2025, restricted the utilization of delandistrogene moxeparvovec to ambulatory patients aged four years and older, and revoked the indication for non-ambulatory patients^55,56. Furthermore, high-dose intrathecal AAV has been associated with dorsal root ganglion toxicity.

Table 4.

Manufacturing and Nonclinical Burdens

Modality class	Best-suited mechanism	Representative approved product	Principal safety obligation
Enzyme replacement	Secreted or lysosomal enzyme deficiency	Imiglucerase (Gaucher type 1)	Lifelong dosing; infusion reactions; anti-drug antibodies
Substrate reduction	Toxic substrate accumulation	Miglustat (Gaucher type 1)	Off-target pathway effects; dose-limiting toxicity
Pharmacological chaperone	Misfolded but partially functional protein	Migalastat (amenable GLA variants)	Eligibility limited to variants designated amenable in the approved prescribing information
Splice modulation (oligonucleotide or small molecule)	Splicing defects and exon skipping	Nusinersen and risdiplam (spinal muscular atrophy)42	Repeated intrathecal dosing for nusinersen, with platelet, coagulation, and urine-protein monitoring
Transcript knockdown (oligonucleotide or RNAi)	Toxic gain-of-function RNA or protein	Tofersen (SOD1-ALS)79,80; patisiran and inotersen (transthyretin amyloidosis)	Boxed Warning and REMS for inotersen; confirmatory clinical-benefit verification for tofersen
mRNA replacement	Transient protein replacement	Investigational as of 8 May 2026	Repeated delivery; immune response to mRNA or lipid nanoparticle
AAV gene replacement	Monogenic loss of function in accessible tissue	Onasemnogene abeparvovec, intravenous and intrathecal77; lunsotogene parvec-cwha (Otarmeni; OTOF-related hearing loss, preserved outer hair cell function and no prior cochlear implant in the treated ear)53	Boxed Warnings for serious liver injury; thrombotic microangiopathy; pre-existing immunity; uncertain durability
Ex vivo lentiviral gene addition	Hematopoietic correction	Atidarsagene autotemcel (metachromatic leukodystrophy)	Boxed Warnings for hematologic malignancy on two products; integration-site monitoring; 15-year follow-up57,59
Genome editing	Correctable or disruptable target	Exagamglogene autotemcel50,51	Off-target editing; large on-target rearrangements; conditioning toxicity; 15-year follow-up
Topical vector therapy	Localized epithelial correction	Beremagene geperpavec (recessive dystrophic epidermolysis bullosa)	Repeated application; caregiver handling precautions; durability requires redosing
Downstream pathway-directed small molecule	Druggable consequences of a genetic lesion	Omaveloxolone (Friedreich ataxia)81; selumetinib and mirdametinib (neurofibromatosis type 1)74–76	No effect on the causal lesion; efficacy claims must be framed at the level of the pathway targeted
Mitochondria-targeted agent	Defined molecular target in mitochondrial dysfunction	Doxecitine and doxribtimine (Kygevvi; TK2 deficiency, symptom onset at 12 years of age or younger)82; elamipretide (Forzinity; Barth syndrome, patients weighing at least 30 kg)83	Confirmatory-trial obligation; liver-transaminase and gastrointestinal monitoring

Note. AAV adeno-associated virus, ALS amyotrophic lateral sclerosis, REMS Risk Evaluation and Mitigation Strategy, RNAi RNA interference, TK2 thymidine kinase 2.

Ex vivo lentiviral platforms require conditioning procedures and pose several risks, including mucositis, infection, infertility, and secondary malignancies. Integrating vectors confer a residual risk of insertional mutagenesis. Elivaldogene autotemcel is associated with a boxed warning for hematologic malignancy; as of July 2025, ten out of sixty-seven trial participants had developed myelodysplastic syndrome or acute myeloid leukemia. An August 2025 label update mandated lifelong monitoring and restricted use to patients without an HLA-matched donor⁵⁷. The European Medicines Agency (EMA) withdrew its authorization in November 2021 at the sponsor’s request, prior to treating any patient under the approval⁵⁸. Lovotibeglogene autotemcel presents a comparable warning, with complete blood counts to be conducted at least semiannually over a fifteen-year period, along with integration-site analysis at six and twelve months⁵⁹. The European Union withdrew the conditional authorization for betibeglogene autotemcel in March 2022 for commercial reasons⁶⁰. Atidarsagene autotemcel employs a different construct and remains authorized in both jurisdictions, with lifelong monitoring required. Genome editing introduces additional concerns, including off-target editing, large on-target rearrangements, vector shedding, transgene immunogenicity, development of neutralizing antibodies that may prevent redosing, conditioning-associated gonadotoxicity, and uncertain durability; these issues necessitate structured monitoring and transparent communication. The section on equity and access addresses the financial implications of single-administration therapy.

Platform development

Platform-based development, in which delivery systems, vectors, manufacturing processes, and regulatory evidence are reused across related products, is an acknowledged methodology to accelerate therapy development for rare diseases. Reuse feasibility depends on the residual disease-specific evidence required, the transferability of immunogenicity and biodistribution data, and manufacturing comparability. Two regulatory frameworks are relevant and should not be conflated. The Platform Technology Designation Program, established by Section 506 K of the Federal Food, Drug, and Cosmetic Act and outlined in the FDA draft guidance of May 2024, permits a sponsor to obtain formal recognition that a technology incorporated in an approved product may be employed in subsequent applications; eligibility hinges upon prior incorporation, preliminary evidence of transferability without negatively impacting quality, manufacturing, or safety, and a reasonable likelihood of achieving significant efficiencies⁶¹. The Plausible Mechanism Framework guidance addresses a different question: producing substantial evidence of effectiveness and safety when clinical data are derived from a limited patient cohort⁹. A designated platform may reduce manufacturing and nonclinical burdens for subsequent personalized products; however, the Plausible Mechanism Framework guidance itself is not a platform designation instrument. (Fig. 3) (Table 4)

Fig. 3.

Translational pipeline from variant discovery to mechanism-based therapy. Variant identification is followed by mechanism characterization and functional validation, then biomarker and endpoint development, and selection of a candidate aligned with the molecular defect. Clinical development employs adaptive, basket, platform, single-arm, registry-based, and N-of-1 designs, supported throughout by registries, natural-history studies, biorepositories, FAIR data sharing, patient organizations, genetic counseling, bioethics, regulatory science, and equitable-access initiatives. AAV adeno-associated virus, ASO antisense oligonucleotide, PRO patient-reported outcome, RMAT regenerative medicine advanced therapy.

Trial design and N-of-1 individualized therapeutics

Why conventional trials are difficult, and what replaces them

Randomized controlled trials are often impractical or ethically constrained in rare diseases. These populations are typically small and geographically dispersed; heterogeneity reduces treatment effects; pediatric onset limits the use of placebo controls; validated outcome measures are often unavailable; and progression rates vary. The FDA guidance recognizes these limitations and discusses the flexibility permitted in study design under the existing statutory requirements for demonstrating substantial evidence of effectiveness and obtaining sufficient information to ensure a drug’s safety⁶². This wording aligns with the governing statutory standards; the guidance does not explicitly state that “standards for safety and efficacy are not diminished.” Viable approaches encompass adaptive and Bayesian trial designs, basket trials organized by shared mechanisms rather than anatomical considerations, platform trials and master protocols, single-arm trials with external controls, delayed-start and randomized-withdrawal protocols, decentralized trials, and registry-based pragmatic studies⁶³.

External controls derived from natural history can substitute for randomization; however, they require meticulous management to account for unmeasured differences in factors, such as genotype, age, sex, ancestry, baseline severity, prior therapy, follow-up duration, and outcome ascertainment bias, which may influence estimates. Depending on the study design and available data, appropriate methodologies may include pre-registered analysis plans, harmonized endpoints and assessment intervals, propensity-score or other adjustments for measured baseline disparities, sensitivity analyses addressing unmeasured confounding, and tipping-point analyses for missing data or deviations from key assumptions. Prospective collection of natural history data aligned with the trial protocol is preferred over retrospective record review. The FDA guidance on rare-disease drug development and externally controlled trials delineates the minimum requirements.

N-of-1 individualized therapeutics

Ultra-rare and patient-specific variants increasingly influence research efforts where interventions are tailored to individual patients or a limited group. Milasen demonstrated that a splice-switching oligonucleotide can be designed, produced, and administered within about one year after molecular diagnosis⁴³, and subsequent initiatives involving personalized oligonucleotides, RNA, and gene editing have followed. A consensus framework delineates essential components: thorough diagnosis, a defined mechanism, preclinical evidence, individualized natural-history baselines, patient-relevant outcomes, dose optimization, safety surveillance, and statistical methodologies suitable for single-patient studies⁶⁴. Strong inference can be achieved through patient-as-own-control designs incorporating treatment withdrawal or washout periods, provided the intervention is reversible, and its effects are expected to diminish, such as with transient RNA-based or other repeatedly administered therapies. Washout procedures are not applicable to genome editing, AAV-mediated gene replacement, or other interventions intended to induce permanent or long-lasting biological modifications. For these therapies, inference should instead rely on prospectively characterized baseline trajectories, biomarker anchoring, calibrated outcome assessments, external or natural-history comparisons where appropriate, and pre-established safety and futility stopping criteria.

Achievements, such as Milasen, are not universally reproducible. Numerous personalized programs fail to reach treatment, experience delays due to manufacturing or toxicology assessments, are halted for safety concerns, yield ambiguous responses, or result in patient mortality due to disease progression before benefits can be evaluated. Negative and inconclusive outcomes are underreported, thereby biasing perceptions of feasibility and effect size. The community has recommended systematic registration, structured reporting, and the routine submission of negative results^64–66.

Regulatory and ethical environment

The draft guidance released by the Food and Drug Administration (FDA) in February 2026 warrants thorough and exact review, as it has been subject to numerous paraphrasing instances that diverge from its original language. Publicly issued for comment on 23 February 2026, with the notice of availability published at 91 FR 9283 on 25 February 2026, under Docket No. FDA-2026-D-1256, it delineates considerations for establishing substantial evidence of effectiveness and safety for personalized therapies based on a plausible mechanism framework. The document specifically addresses genome editing and RNA-based therapies, asserting that the general principles may also be applicable to other individualized therapies. It is particularly relevant in contexts where clinical evidence from a limited number of patients supports safety or efficacy within the target population. As a draft issued under the FDA’s good guidance practices regulation at 21 CFR 10.115, it does not establish binding rights, is non-binding on the FDA or the public, and permits alternative approaches that comply with applicable statutes and regulations. The comment period concluded on 27 April 2026.

Three consequences follow, each more specific than the claims often associated with the document. The framework does not establish a novel approval pathway, nor does it set a new standard of review; statutory requirements for substantial evidence of effectiveness and safety remain in force. The guidance specifies methods for generating such evidence, especially when patient populations are very small. The labeling of the approach as a pathway originated from a Sounding Board article authored by FDA leadership rather than from the guidance itself⁶⁷. Its scope is not explicitly confined to ultra-rare diseases, and the extent to which its principles extend beyond individualized products remains under deliberation pending finalization. Numerical thresholds for evidence are not explicitly defined, and the development community has challenged the associated burdens, advocating for greater flexibility and clearer standards of sufficiency⁶⁸. Previous research anticipated this evolution and promoted frameworks that responsibly encourage innovation in individualized therapeutics⁶⁹.

Ethical obligations encompass addressing therapeutic misconception, securing informed consent in pediatric populations or cases of rapidly progressing diseases, ensuring equitable prioritization when development capacity is limited, clarifying post-treatment responsibilities, and differentiating between research and personalized care. Protective measures include independent ethics review, transparent communication of uncertainties, pre-established stopping criteria, structured safety monitoring, and a clearly defined data-sharing plan^65,66. The moral importance of a single-patient intervention partly depends on whether the knowledge gained is shared in a way that benefits future patients.

Preclinical platforms supporting diagnosis and therapy

Functional validation constitutes a vital stage when sequencing results reveal a variant of uncertain significance. The evaluation process encompasses assessments of enzymatic activity, protein expression levels, transcript abundance, isoform composition, splicing patterns, mitochondrial and lysosomal phenotypes, and tissue-specific parameters, such as cardiomyocyte contractility and neuronal excitability, selected based on gene biology and tissue accessibility³⁰. Patient-derived fibroblasts and induced pluripotent stem cells (iPSCs) facilitate cellular phenotyping and lineage differentiation. iPSC-derived organoids expand modeling into three dimensions. CRISPR-engineered isogenic lines enable variant-specific comparisons. Model organisms remain indispensable, with zebrafish employed for rapid in vivo screening, Drosophila for genetic screening, mouse lines for pharmacological testing and safety assessments, and large-animal models used where murine physiology proves inadequate^70,71.

The evidentiary significance of these findings varies by intended purpose, and conflating different tiers is a common methodological mistake. Evidence sufficient for variant classification under PS3 or BS3 requires a disease-relevant assay with controls, replication, and blinded analysis. To substantiate a therapeutic mechanism, evidence must demonstrate that the intervention rectifies the specific molecular defect within a system that faithfully models the human phenotype. For dose determination, adequate evidence includes pharmacokinetics, biodistribution, and a translatable target-engagement biomarker. Regulatory evaluation also requires additional data, including repeat-dose toxicology, immunogenicity assessments, and, for vector-based products, integration analysis. A functional rescue observed in a patient-derived model generally satisfies the criteria for the initial two tiers; however, it does not establish clinical efficacy nor meet regulatory standards. Preclinical evidence supporting an N-of-1 therapy may be more narrowly focused than that required for broader population products; nevertheless, it must still justify the associated risks to a single patient^9,64. Common pitfalls include inadequate model fidelity, over-reliance on rescue assays, publication bias, and limited follow-up. Integrating data from cellular models, organoids, animal studies, and computational analyses mitigates the risk of a single artifact leading to a misguided therapeutic decision.

Equity, access, and health-system implementation

Equitable access remains the primary gap in rare-disease therapeutics. The underrepresentation of non-European ancestries within reference datasets results in a disproportionate burden of uncertain variants⁷. However, structural barriers extend far beyond reference data, encompassing accredited laboratory capacity, the size and training of the clinical genetics workforce, restrictions on sample export and biobanking, rules concerning data sovereignty, the availability of local ethics review processes, regional manufacturing capacity for cell and gene therapies, supply chains for specialized reagents, and reimbursement frameworks. Rare Diseases International and the Lancet Commission, as well as the International Rare Diseases Research Consortium, have recognized these issues as global health priorities^39,72.

Seven actions would transform that description into a formal agenda: develop ancestry-diverse reference datasets and disseminate them through openly quarriable resources; establish regional diagnostic laboratories with proper accreditation and proficiency testing, instead of routing all testing offshore; share variant-curation capacity via federated expert panels to prevent duplicative effort; link newborn screening condition lists to demonstrated treatment availability within the health system, rather than merely to technical detectability; implement cross-border sample and data governance mechanisms to facilitate federated analysis while preserving local custodianship; employ differential pricing or outcomes-based contracting to ensure that prices correspond with local ability to pay and proven benefits; and allocate funding for long-term surveillance capabilities in all countries where a single-administration therapy is provided, recognizing that a 15-year follow-up mandate is meaningless without the infrastructure to support it.

Newborn genomic screening could accelerate the diagnostic process; however, it introduces considerations absent from existing heel-prick programs. The Genomics England Generation Study, which involves sequencing up to 100,000 newborns and evaluating screening for more than 200 treatable conditions, exemplifies current initiatives⁷³. Critical issues include the boundary between research and clinical service, parental consent, data storage governance, management of uncertain and secondary findings, the implications of false positives, criteria for including conditions, handling of carrier status, equity of access, reanalysis, re-consent upon reaching adulthood, and data protection. Programs should be evaluated based on actionability, false-positive rates, capacity for referral and treatment, equity of outcomes, psychological impact, and sustainability, rather than enrollment alone.

Partnership with patients and families is fundamental, not auxiliary, and should include patient-led registries, preference studies, caregiver-burden assessments, patient-developed outcome measures, and clearly defined roles in governance, protocol development, endpoint selection, recruitment, and dissemination. High therapy costs strain reimbursement systems, and uncertainty about durability complicates outcomes-based contracting; health-technology assessment agencies use varying frameworks, and applying standard cost-effectiveness thresholds without modification creates barriers. The withdrawal of valoctocogene roxaparvovec demonstrates that approval does not guarantee sustained availability⁵⁴. Even without disease-modifying therapy, a confirmed molecular diagnosis facilitates surveillance, prevents ineffective empirical treatments, supports reproductive counseling, validates family experiences, and provides access to registries, clinical trials, and social and educational support. These non-therapeutic benefits are essential when communicating diagnostic value to payers and policymakers.

Outlook

Initial genome sequencing will further expand into diagnostic pathways for selected pediatric, neonatal, neurodevelopmental, metabolic, immunologic, and multisystem indications, with the mechanism-specific exceptions described above still requiring dedicated assays. Advances in curation, structural-variant detection, and noncoding interpretation will facilitate periodic reanalysis^6,8. Long-read sequencing and multi-omics are expected to become increasingly essential problem-solving methodologies for specific unresolved cases; however, factors such as cost, tissue requirements, analytical maturity, reimbursement, and limited validation across diverse settings will influence their incorporation into routine pathways. Artificial intelligence, validated across various cohorts and evaluated for bias, will progressively assist in phenotype coding, variant interpretation, and case identification, in accordance with the safeguards previously noted. Newborn genomic screening will remain subject to debate regarding policy and ethical considerations⁷³.

Rare diseases are individually uncommon but collectively widespread, and genomic medicine has significantly enhanced the capacity to determine molecular etiology without rendering molecular diagnosis sufficient. Effective patient care necessitates structured phenotyping, expert variant interpretation, utilization of multi-omics for unresolved cases, functional validation, context-specific biomarkers, natural history data, registries, and interoperable data sharing. Additionally, preclinical models, mechanism-based therapies, trial designs tailored to small populations, transparent adverse event monitoring, and long-term surveillance are essential. Finalizing the Plausible Mechanism Framework guidance will influence the pace and direction of developing personalized therapies⁹. Furthermore, federated infrastructures built on common data elements, ORPHAcodes, the Human Phenotype Ontology, GA4GH standards, and FAIR principles will determine the reusability of resulting evidence^35,38. Progress depends upon aligning scientific advances with ethical practices, engaging patients as partners, securing regulatory clarity, and closing the access gap. The pace and equity of this progress will be shaped by deliberate and strategic choices.

Limitations

The literature cutoff date is established as 8 May 2026, with regulatory statuses subject to re-verification on 14 August 2026. Approvals, safety communications, and guidance are subject to frequent updates; therefore, reconfirm regulatory statements before making clinical or regulatory decisions. Approvals are specific to individual jurisdictions; unless indicated otherwise, statements herein pertain to FDA decisions, with multinational statuses provided in Supplementary Table S2. The review incorporates peer-reviewed literature and regulatory and industry documents of varying evidentiary strength, cited by relevance. The example disorders presented in Tables 1 and 2 illustrate mechanism classes and their diagnostic and therapeutic implications, not an exhaustive catalog; additionally, disorders categorized under a single mechanism class often exhibit features of other classes. Conclusions derived from selected cohorts, typically pediatric, Mendelian, and from well-resourced centers, should not be presumed applicable to ultra-rare disorders with limited evidence. Diagnostic yield depends on phenotype, age, prior testing, platform, family structure, and pipeline. Platforms utilizing episignatures, multi-omics, and AI-assisted technologies vary in their stages of clinical maturity. Inferences from single-arm trials and external controls are inherently less robust than those from randomized controlled trials. This document constitutes a narrative review rather than a systematic review; the selection of sources is intentional as previously described. The author’s interpretive emphasis in the sections on therapeutics and outlooks should be understood in conjunction with the declared competing interests below.

Author contributions

S.K.N. conceived, drafted, and revised the manuscript and is responsible for the final content.

Peer review

Peer review information

Communications Medicine thanks all reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

No datasets were generated or analyzed for this Review. All sources are identified in the reference list.

Competing interests

S.K.N. declares the following interests. Non-financial: he has served in advisory roles with the U.S. Food and Drug Administration, the European Medicines Agency, and the Medicines and Healthcare Products Regulatory Agency; these agencies did not review, endorse, or contribute to this manuscript. Financial: he holds equity interests in RNA Therapeutics, DEI Biopharma, BioRx, and Abyolo, companies active in therapeutic modality classes discussed in this review. No specific product of these entities is evaluated or endorsed herein. The author acknowledges that these interests may constitute a class-level competing interest with respect to the therapeutic modalities discussed.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s43856-026-01962-3.