SET-Related Intellectual Developmental Disorder

Autosomal dominant intellectual developmental disorder 58 (MRD58), 9q34.11 deletions including SET, 9q34.11 microduplications including SET and differential diagnosis with SETBP1 disorders

Literature and regulatory information updated: 28 October 2026.

0. Reference Summary Table

The scope is the monogenic form caused by SET loss of function and the dosage alterations of the 9q34.11 region that include the gene.

1. Introduction

1.1. From fusion oncogene to intellectual disability gene

The SET gene was identified in 1992 by von Lindern and colleagues as a fusion partner of the CAN gene (now NUP214) in a patient with acute undifferentiated leukaemia; the set-can fusion was described by analogy with the dek-can fusion of the t(6;9)(p23;q34) translocation of acute myeloid leukaemia. The work characterised 2.0 and 2.7 kb transcripts resulting from alternative polyadenylation, a predicted 32 kDa protein with homology to the yeast nucleosome assembly protein NAP-I and a long acidic tail, and localised the gene to 9q34 [26]. In 2001, Seo and colleagues purified a human cellular complex that inhibits histone acetyltransferases, the INHAT complex (inhibitor of histone acetyltransferases), and showed that the SET oncoprotein is one of its components [27]. The name of the gene is unrelated to the SET domain of histone methyltransferases (acronym for Suppressor of variegation, Enhancer of zeste and Trithorax); one of the founding clinical articles wrongly expands the symbol SET in this way, and the confusion recurs in literature searches that mix the SET gene with the SETD genes [5]. The OMIM gene entry repeats this expansion when summarising the 1992 article, which predates the characterisation of the SET domain that the entry itself attributes to Tschiersch and colleagues (1994) [9]. According to the same summary, the predicted SET-CAN chimeric protein was about 155 kDa and retained much of the acidic tail of SET, and fluorescence in situ hybridisation placed SET at 9q34, centromeric to ABL1; the t(6;9), with which the fusion was compared, defines an early-onset, poor-prognosis subtype of acute myeloid leukaemia [9, 26].

1.2. Discovery of the association with neurodevelopmental disorders

The first link between SET and intellectual disability (ID) emerged from an exome sequencing study of 41 trios with moderate or severe ID, carried out in Quebec by Hamdan and colleagues in 2014. Among 81 de novo variants (1.98 per proband), the authors found a de novo deletion creating a stop codon in SET in a 12-year-old boy with moderate ID and congenital microcephaly, without other distinctive features, and proposed SET as a candidate gene [28]. In 2017, the analysis of de novo variants from the Deciphering Developmental Disorders (DDD) study, in the United Kingdom and Ireland, placed SET among the genes with a significant excess of de novo variants in developmental disorders [5, 29].

The clinical entity was delineated in 2018 by two almost simultaneous publications. Stevens and colleagues described four patients with de novo variants and an affected mother and son sharing a frameshift variant (the mother's had arisen de novo), and noted that loss-of-function variants in the general population are exceptionally rare and, when present, affect only one of the gene's transcripts, whereas the patients' variants affect all transcripts [12]. In the OMIM summary, the five children and the mother had developmental delay and ID with language delay and impairment; some had motor delay or poor coordination, but all walked; dysmorphic features were mild and variable (facial asymmetry, high palate, dental crowding, hypertelorism, plagiocephaly, small head circumference, large ears, broad nasal bridge, thick lips and wide mouth), together with hypotonia and joint hypermobility; one patient had two seizures, and brain imaging, performed in a few, was normal. No functional studies of the variants or of patient cells were performed, and loss of function was inferred from the variant type [12, 14]. Richardson and colleagues described three patients from the DDD study with de novo frameshift variants and reviewed the Hamdan case, concluding that the phenotype combined developmental delay, moderate ID and non-specific facial dysmorphic features [5]. Following this work, OMIM assigned phenotype number #618106 to autosomal dominant intellectual developmental disorder 58 (MRD58); the entry was created on 29 August 2018 on the basis of the Stevens series and the DDD and Hamdan cases, and retains the historical designation «mental retardation, autosomal dominant 58» as an alternative title [1, 2, 14]. A 2026 review states that Hamdan and colleagues «designated» MRD58; the 2014 article only proposed the candidate gene, and the designation postdates the 2018 reports [21, 28].

1.3. Consolidation: series from 2022 to 2026

Between 2020 and 2023, isolated cases appeared: a 2.0 Mb deletion including SET and SPTAN1 but not STXBP1 [3], two Chinese patients with splicing and start-loss variants studied by transcriptomics [30], a missense variant coexisting with a de novo STXBP1 variant in a female infant with epilepsy [31] and variants included in large diagnostic series [32, 33]. In 2026, five studies were published that broadened the description of the disease: the largest series, by Shi and colleagues, with 23 previously unreported individuals, computer-assisted facial phenotyping and a DNA methylation episignature [1]; the smallest 9q34.11 deletion associated with a neurodevelopmental disorder, of 326 kb, with fibroblast proteomics and 28 years of clinical follow-up [4]; two Japanese reports with a synthesis of 16 patients [2]; two Chinese reports with a systematic review [21]; and the description of 14 patients with 9q34.11 microduplications encompassing SET, which raises the hypothesis of sensitivity to triplication of the gene [6].

1.4. Definition of the entities within the scope

SET-related neurodevelopmental disorder (MRD58) is defined by the presence of a heterozygous pathogenic or likely pathogenic variant in SET, as a rule a loss-of-function variant, in an individual with global developmental delay or ID. ClinGen classifies the relationship between SET and ID as definitive and records a note from the expert panel: the gene was initially associated with non-syndromic ID, but several patients described have syndromic features, so the curation used the general term «intellectual disability» [10]. Orphanet does not assign the entity a code of its own and includes it in group ORPHA:178469 of autosomal dominant non-syndromic ID [15]. The 9q34.11 deletions that remove SET without neighbouring genes with a phenotype of their own produce the same picture and are covered in this article; the larger deletions that include STXBP1 or SPTAN1 are contiguous gene syndromes in which epilepsy depends mainly on those genes [3, 4]. The 9q34.11 microduplications encompassing SET constitute a distinct entity, described in sections 2.8, 3.12 and 5.8 and in the differential diagnosis, whose causal relationship with SET triplication has not yet been accepted by ClinGen [6, 10]. The somatic SET-NUP214 fusion of T-cell acute lymphoblastic leukaemia is an acquired tumour event, unrelated to the constitutional disease, and is mentioned only to avoid confusion in the interpretation of results [34, 35].

2. Molecular Genetics and Pathophysiology

2.1. The SET gene, its transcripts and the numbering problem

SET lies at 9q34.11 and encodes a ubiquitously expressed nuclear phosphoprotein, also present in the endoplasmic reticulum [5]. The gene has several transcripts that differ in the first exon and therefore at the amino terminus of the protein. The MANE Select transcript (Matched Annotation from NCBI and EMBL-EBI) is NM_003011.4, which encodes isoform 2, also called TAF-Iβ, with 277 amino acids (NP_003002.2). The first clinical reports (Hamdan 2014, Stevens 2018, Richardson 2018), as well as Pan 2023, Mai 2026 and Condell 2026, used NM_001122821 (versions .1 or .2), which encodes isoform 1, TAF-Iα, with 290 amino acids (NP_001116293.1); Condell and colleagues justify the choice as being the longest clinically relevant transcript and the one used in previous reports, and ClinVar, Nishi 2026 and Shi 2026 use MANE [1, 2, 4, 11]. There are also transcripts NM_001248000.2, NM_001248001.2 and NM_001374326.1 [11]. In the shared exons, NM_001122821 numbering exceeds MANE numbering by 39 nucleotides and 13 amino acids; this difference was confirmed variant by variant in VariantValidator and should not be applied by manual calculation, because variants in exon 1 unique to isoform 1 have no corresponding position in MANE [11]. The recurrent MANE p.(Arg44LeufsTer10), for example, corresponds to p.(Arg57LeufsTer10) on NM_001122821 [1, 11].

Constraint metrics confirm intolerance to loss of function: pLI of 1 and LOEUF of 0.17 in gnomAD v4.0 (transcript ENST00000322030) and a DECIPHER haploinsufficiency index (%HI) of 2.03 [10]. In 2018, gnomAD contained only two loss-of-function variants in SET, a splice donor variant (c.112+1G>A) and a single-base deletion (c.112delG), both on NM_001122821 and in the region of exon 1 unique to isoform 1, each with a single allele count [5, 11]. Complete inactivation of Set in the mouse is lethal between embryonic days 11.5 and 12.5 [1].

2.2. Protein structure and functions

The SET protein forms dimers. Each subunit has an amino-terminal region (residues 1–23 of isoform 2), a backbone helix that mediates dimerisation (24–76), an earmuff domain (approximately 79–225), whose underside binds histones and double-stranded DNA, and a carboxy-terminal acidic domain (237–290 in isoform 1 numbering) required for inhibition of histone acetyltransferases and for stimulation of DNA replication; this architecture is similar to that of the NAP-1 protein [1]. A 2026 review describes the domains with a different nomenclature (amino-terminal domain, central acidic domain responsible for protein phosphatase 2A inhibition and carboxy-terminal INHAT domain) [21]; the structural division of Shi and colleagues is based on crystallography (PDB 7MTO and 2E50) and is the one adopted here [1].

The known functions explain the classification of SET as a chromatin regulator. As a component of the INHAT complex, SET binds histones, especially histone H4, and masks the lysines that would be acetylated by the acetyltransferases EP300, CREBBP and PCAF, silencing transcription dependent on these enzymes; it also prevents active DNA demethylation [5, 27]. Isoforms 1 and 2 inhibit protein phosphatase 2A (PP2A). SET acts as a histone chaperone in nucleosome assembly, blocks cyclin B-CDK1 in mitosis, forms with prothymosin-α a complex that decondenses chromatin and interacts with APBB1, CHTOP, SETBP1, SGO1 and MCPH1 [5]. Isoform 2 has an antiapoptotic role: it is part of the SET complex (with ANP32A, APEX1, HMGB2, NME1 and TREX1) and, when cleaved by granzyme A, releases the DNase NME1 [5]. In the nervous system, experimental reduction of SET/TAF-Iβ induces neuronal differentiation, identifying it as a negative regulator of differentiation, and the gene takes part in the neural cell cycle, neurite outgrowth and neuronal survival [1, 5]. Reduction of SET reproduces in cells the premature chromosome condensation seen with loss of MCPH1, but this cellular phenotype has not been investigated in patients [5].

The OMIM summary adds four experimental observations. The interaction with SETBP1, identified by yeast two-hybrid screening with the 277 amino-terminal amino acids of SET as bait, depends on residues 182–223 of SET and residues 1238–1434 of SETBP1, and was confirmed with the endogenous protein in an osteosarcoma cell line; it is the molecular basis of the SET–SETBP1 axis discussed in the differential diagnosis. SET expression is high in proliferating cells, in the developing kidney and in Wilms tumour, and decreases with quiescence and differentiation. In the mouse brain, kainic acid activates lysosomal asparaginyl endopeptidase (legumain), which degrades Set and is followed by DNA nicking and neuronal death; the PIKE-L protein protects Set from this cleavage, and mice lacking the endopeptidase do not suffer this injury after induced seizures or stroke. Finally, the acidic domain of SET binds the non-acetylated lysine-rich regions of p53, histone H3, Ku70 and FOXO1, and represses p53; acetylation of p53 after DNA damage abolishes the binding, which makes SET a reader of lysine acetylation and not merely an inhibitor of acetyltransferases [9]. None of these observations has been studied in cells from patients with MRD58.

2.3. Disease mechanism: haploinsufficiency

The accepted mechanism is haploinsufficiency. ClinGen gives it a score of 3 (sufficient evidence) [10], and the supporting data are convergent. First, the great majority of patient variants are truncating and predicted to undergo nonsense-mediated mRNA decay (NMD); those escaping such decay would remove the earmuff domain and the acidic tail and produce unstable subunits unable to bind histones and DNA [1]. Second, deletions removing SET without neighbouring genes with a phenotype of their own produce the same picture [2, 4]. Third, quantitative proteomics of fibroblasts from the patient with the 326 kb deletion showed reduction of all proteins encoded in the interval, to between 46% and 81% of control levels (five controls), with canonical SET protein reduced to 50%, while proteins encoded by flanking genes outside the deletion (SPTAN1, SH3GLB2, DOLPP1) remained normal [4]. In two Chinese patients, RNA sequencing showed extensive dysregulation of genes of the NF-κB pathway and the neuronal system [30].

Loss of half of the inhibitory activity of the INHAT complex would be expected to translate into histone hyperacetylation. The DNA methylation data of Shi and colleagues point in this direction: the differentially methylated regions in patients coincide, in neural progenitor cells derived from embryonic stem cells and in fetal brain, with chromatin states of active promoters, weak transcription and enhancers, including bivalent enhancers, and also with regions normally repressed by the Polycomb complex, which the authors interpret as extensive remodelling of the chromatin landscape [1]. The opposite effect, excess SET, occurs in Schinzel-Giedion syndrome: gain-of-function SETBP1 variants in the degron prevent the degradation of SETBP1, which binds SET and protects it from ubiquitin-mediated degradation; accumulation of SET causes histone hypoacetylation at enhancers and a reorganisation of distal enhancer usage during neural fate specification [1, 18, 36]. In SETBP1 haploinsufficiency disorder (MRD29), less stabilisation of SET is expected, on the contrary, which brings MRD29 mechanistically close to MRD58 [1]. The hypothesis that imbalance of SET dosage, in either direction, disturbs cis regulation of neuronal transcription is the authors' own and has not been demonstrated directly in tissue from patients with MRD58 [1, 36].

2.4. Inheritance, penetrance and expressivity

Inheritance is autosomal dominant. In the Shi series, inheritance was studied in 22 individuals and the variant arose de novo in 21 (95%); the exception was the recurrent p.(Arg44LeufsTer10), inherited from a probably affected father about whom little information was available [1]. In the Stevens series, a woman with the de novo variant transmitted it to her son, both affected [4, 12]. In two published patients (Pan 2023, the patient with start loss, and Brunet 2021) inheritance was not established [21]. The sources do not describe asymptomatic carriers of loss-of-function variants; penetrance appears complete for developmental delay, but expressivity varies between borderline intellectual functioning and severe ID, and between the absence and presence of seizures, ocular problems or sleep disorder [1, 4]. No consistent genotype–phenotype correlation has been identified; the authors of the largest series note that the total number of patients remains insufficient to detect subtle associations [1]. The only trend described is greater severity of ID in the two deletions of the region (Shono 2022 and Condell 2026) compared with point variants, with the caveat that these deletions remove other genes [4].

2.5. Variant spectrum and normalisation

On 28 October 2026, ClinVar recorded 144 single-gene variants in SET: 22 pathogenic, 4 pathogenic or likely pathogenic, 17 likely pathogenic, 54 of uncertain significance (VUS), 24 likely benign, 4 benign, 1 with conflicting classifications and 18 unclassified or in other categories [37]. Pathogenic variants are distributed across the whole protein and include frameshift, nonsense, splicing, start-loss, rare missense variants and an in-frame deletion [1, 37]. The variant p.(Arg44LeufsTer10) is recurrent: it was described by Stevens (mother and son), by Richardson (two patients), by Nishi and in 10 of the 26 individuals of the Shi series [1, 2, 5]. The 4-nucleotide deletion occurs in a repeated sequence; VariantValidator normalises it to NM_003011.4:c.128_131del, whereas the title of ClinVar record VCV000521454 presents it as c.130_133del; this is the same genomic change (NC_000009.12:g.128691224_128691227del) [11].

The following table brings together published variants with clinical data, with the original notation, the conversion to MANE and the genomic position, all obtained from VariantValidator (GRCh38) [11].

The Shi series was published directly on MANE and includes 17 different variants in 26 individuals: ten frameshift — p.(Ser9LysfsTer33), p.(Lys26ArgfsTer7), p.(Lys26AsnfsTer16), p.(Asp36LysfsTer6), p.(Arg44LeufsTer10), p.(Gln69SerfsTer33), p.(Glu73AspfsTer2), p.(Val110SerfsTer5), p.(Asn215LysfsTer13) and p.(Glu243AspfsTer2) —, one nonsense, p.(Trp213Ter), one splicing, c.493-3_504del, one MANE start-loss, c.1A>G p.(Met1?), two missense, p.(Arg64Cys) and p.(His105Arg), one in-frame deletion, p.(Asp36del), and a deletion of about 400 kb including SET and SPTAN1 (NC_000009.12:g.128515531_128922357del) [1]. All were absent from gnomAD v4.1.0, except c.1A>G, observed 3 times in 1,305,368 alleles [1]. Under the ACMG/AMP standards, 14 were classified as pathogenic and 1 as likely pathogenic; p.(Asp36del) and p.(Arg64Cys), considered causal by the referring centres, were reclassified as VUS [1, 38]. Structural modelling predicted a more marked effect for p.(His105Arg), which alters a residue of the earmuff domain in contact with histones and DNA, and moderate effects for the two VUS, both in the dimerisation helix [1]. MANE residue His105 is also the one affected by the Stevens missense p.(His105Tyr) [11, 37].

OMIM lists five of these variants as allelic variants (600960.0001 to 600960.0005), all in NM_001122821.1 numbering, and reading them requires two corrections. The Hamdan variant, c.699_701del, is described there as frameshift, but it is an in-frame three-nucleotide deletion that directly creates a stop codon, p.(Tyr233Ter), as VariantValidator confirms; and the variant p.(His118Tyr) is attached to transcript NM_001122321.1, a typographical error for NM_001122821.1 [9, 11]. The same entry places the two Stevens missense variants at a conserved residue of the NAP-like domain, responsible for binding to histones and double-stranded DNA, and records that the recurrent deletion c.167_170del had been observed de novo in two DDD patients before the description of the Stevens family [9].

2.6. Variants in exon 1 unique to isoform 1

Two published notations require care, because they fall in exon 1, which exists only in isoform 1. The first is NM_001122821.1:c.3G>C, p.(Met1?) (in the current version of the transcript, NM_001122821.2:c.3G>C, NC_000009.12:g.128683898G>C), of Pan and colleagues: it abolishes the isoform 1 start codon but has no MANE equivalent, whose translation starts in a different first exon; isoform 2 remains, in principle, intact [11, 30]. This contrasts with the observation of Stevens that pathogenic loss-of-function variants affect all transcripts, whereas those observed in the population affect only one; the Pan variant should therefore be interpreted with caution, even though the patient has moderate ID [12, 30]. The MANE start loss, c.1A>G, described by Shi, is a different variant, in a different exon [1, 11].

The second is the variant of Marinakis and colleagues, cited by Condell as c.99_102del p.(Arg35LeufsTer10) [4, 32]. On NM_001122821.2, c.99_102del would be a deletion in exon 1 unique to isoform 1, with effect p.(Pro34ArgfsTer21) and no MANE equivalent. The protein description p.(Arg35LeufsTer10), however, corresponds on transcript NM_001248000.2 to c.101_104del, which is the same genomic change as the recurrent variant (g.128691224_128691227del); VariantValidator automatically remaps NM_001248000.2:c.99_102del to c.101_104del [11]. The most likely interpretation is that the Greek patient has the recurrent variant described on another transcript; the transcript used in the original article is not stated in the sources consulted, so the conversion remains a deduction based on the agreement of the protein description [11, 32]. The example shows why the laboratory report should state the transcript with its version.

2.7. 9q34.11 deletions including SET

Deletions of the region allow the contribution of SET to be delineated. In 2018, DECIPHER contained 10 cases with a loss including SET, from 703 kb to 4.13 Mb, five de novo, three of unknown inheritance and two paternal; six of eight with a described phenotype had ID and none had seizures; the only other monoallelic DDG2P gene common to all was SPTAN1 [5]. Shono and colleagues described a 2.0 Mb deletion with SET and SPTAN1 but without STXBP1; among the loss-of-function-intolerant genes in the interval, only SET and SPTAN1 had a %HI below 10 [3]. In the Shi series, the 400 kb deletion with SET and SPTAN1 occurred in an individual with epilepsy, and the authors do not exclude a contribution from SPTAN1, associated with autosomal dominant epileptic encephalopathy [1]. Nishi described a 1.04 Mb deletion, arr[GRCh37]9q34.11(131455365_132494818)x1, which includes SET and the OMIM genes DOLK, CRAT, PTPA, LRRC8A and NUP188 [2].

The deletion of Condell and colleagues, of about 326 kb and de novo, extends from DYNC2I2 to NUP188 (13 genes) and was confirmed by single-nucleotide polymorphism (SNP) microarray and by nanopore long-read sequencing; it includes neither STXBP1 nor SPTAN1 and is the smallest 9q34.11 deletion associated with a neurodevelopmental disorder, defining a critical region centred on SET [4]. Under the ACMG/ClinGen CNV standards, it scored 1.15 points (pathogenicity threshold 0.99): 1A 0 points, 2A 1.00 point for complete overlap of a gene with established haploinsufficiency, 3A 0 points for the number of genes and 5A 0.15 points for de novo occurrence with a compatible but non-specific phenotype [4, 13]. Long-read sequencing also showed a paternal missense variant in NUP188 in cis with the deletion; immunofluorescence showed no reduction of NUP188, and the variant was not considered contributory [4].

Larger 9q33.3-q34.11 deletions, which include STXBP1, produce a different phenotype, often with epileptic encephalopathy. Ehret and colleagues, cited by Condell, suggested that STXBP1 is not the only causal gene in these deletions [4, 39]. The GeneReviews review of STXBP1 brings together 20 individuals with contiguous deletions of up to 4 Mb and up to 92 genes, in whom, besides epilepsy, isolated findings appeared such as absent thumbnails, cleft lip and palate, ventricular septal defect, renal dysplasia and microcephaly [19].

2.8. Dosage gain: 9q34.11 microduplication including SET

De Falco and colleagues, in a consortium of centres in Italy, France, Poland and Australia, started from an 11-year-old girl with speech delay, ID and behavioural changes carrying a de novo 9q34.11 microduplication including SET and SPTAN1, and gathered another 13 patients with overlapping duplications, 14 in total, published in 2026 [6]. In addition to the neurodevelopmental disorder, the patients shared recurrent facial dysmorphic features, mainly midface hypoplasia and thin lips, and the smallest region of overlap of the duplications contained SET, which led the authors to propose that SET triplosensitivity contributes to the phenotype and to describe the group as a 9q34.11 microduplication syndrome [6]. Shi and colleagues cite this work as an argument that SET is dosage-sensitive in both directions [1]. The sources consulted do not specify, patient by patient, the size of the duplications, the number of genes included, the parental origin outside the index case or the existence of asymptomatic carrier relatives, which are the data on which any estimate of penetrance depends.

Biological plausibility rests on the function of SET and on the SETBP1–SET axis. SET is an endogenous inhibitor of the phosphatase PP2A and a component of the inhibitor of histone acetyltransferases complex (INHAT), which masks histone tails and reduces their acetylation (sections 2.2 and 2.3). In Schinzel-Giedion syndrome, gain-of-function SETBP1 variants prevent degradation of this protein, increase the amount of SET, reduce histone acetylation and compact chromatin at enhancers, a human model in which excess SET accompanies a severe neurodevelopmental disorder [18, 36]. The extrapolation has two limits: in that model the increase in SET is indirect and adds to other effects of SETBP1, and in a constitutional duplication the expected increase is about half the gene dose, with no published measurement of the protein in patients. The severity of Schinzel-Giedion syndrome therefore does not serve as a reference for the microduplication.

Neighbouring genes make attribution of the phenotype to SET difficult. The duplication in the index case includes SPTAN1, and the larger duplications of the region extend to other genes, such as DNM1 and GOLGA2 [6, 22]. In ClinGen, SET, SPTAN1 and STXBP1 have a haploinsufficiency score of 3 and a triplosensitivity score of 0 (no evidence); the last assessment of SET dates from 22 January 2025 and that of SPTAN1 from 24 September 2025, both before the online publication of the De Falco series in November 2025, and the list of curated regions contains no 9q34.11 interval [10, 20]. Pathogenic SPTAN1 variants act mainly through a dominant-negative effect, and no phenotype due to copy gain of the gene has been described [3].

Population data limit interpretation in both directions. In the gnomAD v4 structural variant dataset, which brings together genomes from about 63,000 individuals, there are two duplications encompassing the whole of SET: one of about 603 kb (GRCh38 chr9:128105846-128709292), with 19 coding genes, including SET, SPTAN1, DNM1 and GOLGA2, observed in 2 of 126,040 alleles, and another of about 47 kb (chr9:128658552-128706000), containing only SET and PKN3, observed in 1 of 123,770 alleles [22]. The frequency is low, of the order of 2 per 100,000 alleles, and does not correspond to common population variation; but the presence of a duplication confined to SET in a population database that excludes, as far as possible, severe paediatric disease shows that SET gain is not incompatible with the absence of severe disease and is compatible with incomplete penetrance or mild expressivity. gnomAD has no cognitive data on participants, and short-read sequencing does not determine whether these copies are in tandem or inserted elsewhere in the genome [13, 22].

The divergence between the De Falco proposal and the ClinGen score of 0 persists in November 2026. The clinical association rests on 14 patients gathered from an index case, in a population referred for neurodevelopmental disorder, in which the phenotype is non-specific and frequent, without a case-control study or published segregation data by family. A duplication including SET is classified under the ACMG/ClinGen standards for copy-number gains, without criterion 2A (complete overlap of a gene or region with established triplosensitivity), scoring the case evidence from the literature and the inheritance in the proband, and with reassessment when ClinGen reviews the gene; the application of the criteria is detailed in section 5.8 [6, 10, 13].

The 9q34.11 microduplication is distinguished from two neighbouring entities. Terminal 9q34.11→qter duplications, as a rule due to a derivative chromosome from a translocation, also encompass SET, but include hundreds of genes and correspond to distal 9q duplication (ORPHA:96101), defined by Orphanet by psychomotor and speech delay, ID, hypotonia, a long slender habitus, craniofacial dysmorphism (dolichocephaly or microcephaly, facial asymmetry, narrow palpebral fissures, deep-set eyes, strabismus, microstomia, retrognathia), arachnodactyly, camptodactyly and flexion contractures [7]. A published example is the de novo duplication of about 9.7 Mb, arr[GRCh37] 9q34.11q34.3(131348076_141019088)x3, with 246 genes, resulting from a der(15)t(9;15)(q34;p11) chromosome, in a Mexican boy with aphallia, urethrorectal fistula, hypoplasia of the corpus callosum, hypotonia and mild ID; the interval contains SET, but the phenotype cannot be attributed to a single gene [16]. 9q34.2-q34.3 duplications have been associated with congenital heart disease, including tetralogy of Fallot with absent pulmonary valve [40]. 9q34.3 microduplications encompassing EHMT1, the Kleefstra syndrome gene, do not include SET and cause a mild neurodevelopmental disorder, without constant dysmorphic features, with a DNA methylation profile of its own and a weak signal [8, 41].

2.9. Allelic diseases and somatic events

No constitutional allelic disorders of SET other than MRD58 have been described. In leukaemia, SET takes part in somatic fusions: the 1992 set-can fusion [26] and the SET-NUP214 fusion, generated by a recurrent cryptic deletion del(9)(q34.11q34.13) in paediatric T-cell acute lymphoblastic leukaemia, in which the chimeric protein binds HOXA promoters with CRM1 and DOT1L and activates these genes [34]. The fusion was also described in an early T-cell precursor acute lymphoblastic leukaemia in a 44-year-old woman, with relapse less than one month after remission [42]. Orphanet records the association of SET with T-cell precursor acute lymphoblastic leukaemia as «part of a fusion gene» [35]. These events are acquired in tumour cells and do not indicate cancer predisposition in constitutional carriers; the clinical sources on MRD58 describe no neoplasms in patients [1, 4].

3. Clinical Spectrum and Manifestations by System

3.1. The available cohorts

The natural history rests on small series and isolated reports, all of patients not treated with specific therapy, because none exists; the data therefore refer to the course under rehabilitation and usual care. The series of Shi and colleagues (2026) brings together 23 individuals with pathogenic or likely pathogenic variants recruited through GeneMatcher and a network of centres in the Netherlands, the United States, Spain, Switzerland, France, the United Kingdom and Canada; it includes 10 females and 13 males and a fetus from a pregnancy terminated at 26 weeks, and excludes the individual with the 400 kb deletion and the two with VUS. In the 22 living individuals, the age at last assessment ranged from 19 months to 19 years (mean 8.3 years; median 8 years; interquartile range 3–12 years). The authors compared the frequencies with 15 individuals from the literature [1]. The table of Richardson and colleagues describes four patients aged 10 to 17 years in detail [5]; Nishi and colleagues summarise 16 patients [2] and Mai and colleagues another 16 [21], with overlap between these syntheses. The frequencies that follow depend on the denominators, which vary because not all findings were assessed in all individuals.

Source: Shi and colleagues, table 2 [1].

3.2. Prenatal and neonatal period

Most individuals were born after an uneventful spontaneous pregnancy, at term, with an uncomplicated neonatal period and normal anthropometric measurements; there was intrauterine growth restriction in 1 of 21 [1]. In the Richardson patients, birth weight ranged between the 5th and 79th centiles and head circumference between the 1st centile (the Hamdan case, with congenital microcephaly) and the 96th centile [5]. In one patient, fetal ultrasound showed unilateral renal agenesis [5]. The Shi series includes a 26-week fetus from a terminated pregnancy, with limited information [1].

3.3. Psychomotor development and cognition

All individuals with information had global developmental delay in the first year of life. Cognitive outcome was learning difficulties or borderline intellectual functioning in 4 of 17 (23.5%) and ID, almost always mild, in 9 of 17 (52.9%) in the Shi series; in the previous literature ID was reported in 12 of 13 (92.3%), mostly moderate [1, 5]. In 7 individuals with formal assessment, the intelligence quotient ranged from 35 to 89 (mean 68.7; median 70; interquartile range 64–84) [1]. The four Richardson patients had moderate ID and attended special school, one of them also mainstream education [5]. In the Nishi patients, the developmental quotient on the Kyoto scale was 66 at 7 years in the carrier of the recurrent variant (mild ID, with the postural-motor domain at 43) and 44 at 3 years in the patient with the 1.04 Mb deletion (moderate ID) [2]. The two Chinese patients of Mai had severe global delay at 2 and 3 years [21]. Severe ID has been described only in the two deletions (Shono and Condell) [4]. No regression was described with point variants in the Shi series [1]; the Condell patient, with a deletion, had motor regression in childhood (section 3.11) [4].

3.4. Speech and language

Speech delay is constant and is, for several teams, the most persistent deficit. In the Shi series, the mean age at first words was 25.5 months (median 24; interquartile range 18–31.5 months; n = 12); 20 of 22 individuals (90.9%) acquired speech, and the two without speech were under 30 months at last assessment; 13 had language delay or impairment defined as first words at 16 months or later, or speech limited to words or short phrases inappropriate for age [1]. In the Richardson patients, first words appeared between 24 and 48 months; the Hamdan case had 20 words at 9 years [5]. Mai and colleagues recorded speech and language delay in 16 of 16 patients and suggest that language is disproportionately affected relative to motor development, a hypothesis that, by their own account, requires more data [21]. Condell and colleagues observe that absent or severely delayed speech in the early years appears to be core, that some patients retain or recover short phrases with repertoires of 2 to 20 words and that speech may improve into adulthood with speech therapy; the mother described by Stevens had speech delay as a child and, at 49 years, good speech and the ability to read short segments of simple text [4].

3.5. Neuromuscular system, motor function and epilepsy

Axial hypotonia was recorded in 14 of 20 individuals (70.0%) in the Shi series, with delayed head control in 7 of 19 (36.8%) [1]; in the Nishi and Mai syntheses the frequency was 5 in 6 and 5 in 16, respectively, which mainly reflects the lack of systematic recording [2, 21]. Independent walking was achieved by 18 of 22 individuals (81.8%), at a mean of 28.4 months (median 29; interquartile range 23.3–34.5 months; n = 16); the four who were not walking were under 30 months [1]. In the Richardson patients, walking appeared between 25 months and 3 years [5], and Nishi notes that many published patients walk between 24 and 36 months [2]. There was spasticity in one individual and hyperreflexia in three; no movement disorders or neuropathy were observed [1].

Epileptic seizures occurred in 4 of 22 individuals (18.2%) in the Shi series, in the form of generalised seizures and staring episodes, and in 1 of 6 from the literature [1]. In the Mai review, 3 of 16 patients (18.75%) had seizures [21]. Mai patient 1, with the in-frame deletion p.(Thr98del), had recurrent afebrile focal seizures from 22 months, controlled with levetiracetam, with occipital slowing while awake and frontocentral epileptiform discharges during sleep [21]. The female infant described by Zhang, with frequent convulsions from 2 months, also had a de novo STXBP1 variant, to which the epilepsy is attributable [21, 31]. Epilepsy is therefore not a core feature; when it is early and severe, it should prompt a search for a second cause or for a deletion including STXBP1 or SPTAN1 [1, 21].

3.6. Brain imaging

Brain magnetic resonance imaging (MRI) was performed in 16 individuals in the Shi series, with non-specific abnormalities in 6 (37.5%): enlarged ventricles, prominent perivascular spaces, thin corpus callosum and possible ischaemic sequelae [1]. In the individual reports, MRI was normal in most [2, 4, 21]; one Richardson patient had left periventricular hyperintensity at 22 months, interpreted as a possible normal variant [5]. There is no characteristic imaging finding.

3.7. Behaviour, sleep and mental health

In the Shi series, 9 of 21 individuals (42.9%) had behavioural changes, including aggression in one and attention-deficit/hyperactivity disorder (ADHD) or hyperactivity in six; sleep disorder was recorded in 9 of 22 (40.9%) [1]. In the Nishi synthesis, 5 of 15 patients had a formal diagnosis of a neurodevelopmental disorder, such as autism spectrum disorder (ASD) or ADHD, 5 of 12 had behavioural difficulties and 5 of 9 had sociable and friendly traits [2]. The Richardson patients illustrate the variability: one without behavioural problems, one with tantrums, hyperphagia and autism, one with mild anxiety and one with attention deficit without hyperactivity; the patient with autism developed schizophrenia at 15 years, an isolated finding [5]. The adult Condell patient has ASD and significant anxiety [4].

3.8. Eyes and vision

Ocular and visual problems are frequent and heterogeneous: 13 of 19 individuals (68.4%) in the Shi series, with refractive errors, astigmatism, anisocoria, optic disc pallor, strabismus, amblyopia and nystagmus [1]. Strabismus was described in two of the three DDD patients, in the Condell patient from 2 years, in one Pan patient and in Mai patient 1, who had congenital strabismus [4, 5, 21]. Mai patient 2 had prolonged P100 latency on visual evoked potentials [21].

3.9. Growth, feeding, skeleton and other organs

Growth is usually normal, but short stature has been described in some patients: the Hamdan case had weight, height and head circumference below −2 standard deviations at 9 years (head circumference −3.9) [5]; one Pan patient had height below the 3rd centile at 11 years [21]; and the Nishi patient with a deletion had progressive short stature (−3.0 standard deviations at 3 years and −3.6 at 3 years and 11 months) without growth hormone deficiency [2]. Feeding problems occurred in 4 of 22 individuals (18.2%) [1]. Recurrent respiratory infections occurred in 5 of 22 (22.7%), frequent otitis in three and mild hearing loss in one [1]; one DDD patient with chronic cough had a right tracheal bronchus [5]. Scoliosis was observed in 2 of 22 (9.1%) [1]. The Mai review found skeletal anomalies in 9 of 16 patients (56.3%) and flat feet in 3 of 11, and proposes them as a diagnostic clue: brachydactyly, syndactyly, clinodactyly, joint laxity and thumb polydactyly [5, 21]. The adult Condell patient has kyphosis and persistent constipation, which the authors put forward as a possible intestinal component to be clarified [4]. Major structural malformations are rare: one ventricular septal defect in the Shi series, one unilateral renal agenesis, one patent foramen ovale [1, 5, 21]. No major anomalies of the limbs, thorax or abdomen have been described [1].

3.10. Dysmorphic features and computer-assisted facial phenotyping

Dysmorphic features were recorded by clinicians in 16 of 22 individuals (72.7%) in the Shi series, but they were mild and non-specific: high anterior hairline (8), abnormal eyebrows, thin, thick, arched or with synophrys (8), epicanthic folds (10), hypertelorism (5), short philtrum (5), mild frontal bossing (4), bulbous nose (4) and broad nasal bridge (4) [1]. The Nishi synthesis, with denominators of 12, records periorbital fullness (12/12), broad nasal bridge and tip (12/12), thick lower lip (11/12), wide mouth (10/12), downslanting palpebral fissures (9/12), broad or prominent forehead (7/12), widely spaced teeth (7/12) and hypertelorism (5/12) [2]. Shono described plagiocephaly, facial asymmetry, a broad high forehead, wide mouth and prominent mandible [3]; the DDD dysmorphologists considered the similarities insufficient for an easily recognisable syndrome, with a wide mouth and thick lower vermilion, broad-based nose and widely spaced teeth as common features [5].

Analysis by GestaltMatcher, a convolutional neural network that compares photographs in a facial phenotype space, confirmed the absence of a gestalt: in the comparison of 9 individuals from the series with 3 published cases, most distances were above the similarity threshold, with a slight shift of the distribution suggesting partial similarity in a small subgroup (seven individuals), which was nonetheless not significant [1]. In the comparison with other chromatin disorders — SETBP1 haploinsufficiency (MRD29), Schinzel-Giedion syndrome, SETBP1 disorder due to variants outside the degron, Menke-Hennekam syndromes 1 and 2, Rubinstein-Taybi syndromes 1 and 2 and Cornelia de Lange syndrome 1 —, patients with SET did not cluster with any of them [1]. Nishi adds that the facial morphology is subtle and possibly age-dependent, which limits the diagnostic value of dysmorphology alone [2].

3.11. Adulthood

Data in adults come from three individuals. The mother described by Stevens had moderate ID, good speech and basic reading at 49 years and had an affected son [4, 12, 21]. The Condell patient, with the 326 kb deletion, was followed up to 28 years: severe ID stable for several years, walking possible but with fatigue (usually less than 200 m at a time), declining hand function, very limited speech but with some phrases, constipation, kyphosis, anxiety, two episodes of loss of consciousness with no identified cause and a normal EEG; she lives in supported residential care [4]. In childhood, this patient had a picture overlapping atypical Rett syndrome, with motor regression, hand stereotypies, bruxism, episodic hyperventilation, small cold hands and feet and reduced response to pain, features not described in other patients with SET and which may depend on the other genes in the deletion [4]. The DDD patient with schizophrenia at 15 years is the only one with psychotic illness described [5].

3.12. 9q34.11 microduplication: clinical picture described

The published phenotype of the 9q34.11 microduplication comes from the 14 patients of De Falco and colleagues [6]. The index case, a girl assessed at 11 years, had speech delay, ID and behavioural changes; the patients as a whole shared a neurodevelopmental disorder and recurrent facial dysmorphic features, in particular midface hypoplasia and thin lips [6]. The speech delay and ID overlap with the MRD58 picture, but the dysmorphic features described differ: in SET loss of function there is no recognisable facial gestalt, not even by computational analysis, whereas in the duplication the authors identified recurrent facial features [1, 6]. The sources consulted do not provide the frequencies of each manifestation, the degree of ID per patient, the occurrence of seizures, associated malformations, growth or the course in adulthood, so the natural history of the microduplication has not been established. Comparison with the population duplications in gnomAD (section 2.8) suggests that clinical expression may be variable and, in some carriers, mild or absent [22].

4. Clinical Recognition and Red Flags

4.1. When to suspect

SET-related disorder has no clinical signs that allow it to be suspected before genetic testing. The face is not recognisable, either on examination by a dysmorphologist or by computational analysis, and there are no biochemical or imaging markers [1, 5]. In the DDD patients, the diagnoses considered before the exome were Williams syndrome (two patients) and Pitt-Hopkins syndrome (one), and the Condell patient was followed for years as atypical Rett syndrome, which illustrates the lack of specificity [4, 5]. For this reason, the authors of the series do not recommend testing targeted at SET and advocate exome or genome sequencing as first line [5, 21]. In practice, the diagnosis emerges as a result of the general genetic investigation of global developmental delay or ID.

There is, however, a set of findings that makes the diagnosis plausible when a SET variant is found: global developmental delay with hypotonia in the first year; disproportionate speech delay, with first words around 2 years and late phrases; walking between 2 and 3 years; progression to mild ID or borderline functioning; ocular problems (strabismus, refractive errors); sleep disorder; and mild dysmorphic features, such as a wide mouth, thick lower vermilion, broad nasal tip, periorbital fullness and widely spaced teeth [1, 2, 5].

4.2. Red flags and signs prompting reassessment

Some findings do not belong to the usual picture and should prompt a search for a second cause or a contiguous deletion. Epilepsy with onset in the neonatal period or the first three months of life, infantile spasms and epileptic encephalopathy suggest a defect of STXBP1 or SPTAN1, through a coexisting point variant or a deletion including these genes [1, 19, 31]. Severe ID, motor regression, hand stereotypies and hyperventilation episodes have been described only in one deletion and should prompt review of the differential diagnosis with Rett syndrome and other genes in the interval [4, 43]. Regression was not observed in individuals with point variants [1]. Severe postnatal microcephaly, present only in the Hamdan case, and progressive short stature justify targeted investigation, as was done for growth hormone deficiency in the Nishi case [2, 5]. Unexplained loss of consciousness in an adult, described once, requires the usual neurological and cardiological assessment, because there are no data attributing it to the disease [4].

4.3. Newborn screening

There is no newborn screening and no marker that would allow it. The disease can be identified only by genetic testing.

5. Diagnosis: Criteria, Testing and Algorithms

5.1. Criteria

There are no published clinical criteria. The diagnosis is established in a proband with global developmental delay or ID and one of the following alterations: a heterozygous pathogenic or likely pathogenic SET variant, classified under the ACMG/AMP standards [38]; or a deletion including SET, classified under the ACMG/ClinGen CNV standards [4, 13]. Truncating variants, canonical splicing variants and whole-gene deletions as a rule meet the PVS1 criterion, because haploinsufficiency is the established mechanism; confirmed de novo occurrence supports PS2 and absence from gnomAD PM2, as in the classifications of Nishi (PVS1, PS2, PS3, PM2) and Mai (PVS1, PS2 and PM2 for the frameshift; PS2, PM4 and PM2 for the in-frame deletion) [2, 21, 38]. Applying PVS1 to variants in exon 1 unique to isoform 1 is not acceptable without proof of an effect on all transcripts (section 2.6) [11, 12]. SET is rated green, with diagnostic-grade confidence, on the Genomics England and Australian Genomics ID panels in PanelApp [4, 44].

5.2. Tests and order of testing

The 2021 American College of Medical Genetics and Genomics (ACMG) guideline, based on the GRADE methodology, strongly recommends that exome or genome sequencing be considered as a first- or second-tier test in paediatric patients with congenital anomalies with onset before 1 year or with developmental delay or ID with onset before 18 years, because it has a higher diagnostic yield than conventional tests and may be more cost-effective when requested early [23]. For SET-related disorder, the tests have the following role:

In practice, faced with a VUS in SET, the available resources are parental testing (a confirmed de novo occurrence with a compatible phenotype weighs in favour of pathogenicity), the methylation episignature, RNA sequencing for splicing variants and reverse phenotyping guided by the findings in sections 3 and 4 [1, 30, 38]. MRI, EEG and biochemical findings neither confirm nor exclude the diagnosis [1].

5.3. DNA methylation episignature

Shi and colleagues compared peripheral blood DNA methylation of 7 individuals with pathogenic SET variants with matched controls and identified 216 differentially methylated CpG probes; hierarchical clustering and multidimensional scaling analysis clearly separated cases and controls [1]. A support vector machine classifier was trained on 75% of the controls and samples from other neurodevelopmental disorders and tested on the remaining 25%; the discovery cases obtained methylation variant pathogenicity scores close to 1, the test samples obtained low scores, and sensitivity and specificity were compared with the clinical EpiSign classifier version 5, which brings together more than 100 episignatures; cross-validation kept all cases clustered with the discovery cohort [1]. The study does not publish numerical values of clinical sensitivity and specificity, and the discovery cohort is small. Applied to the two VUS, p.(Asp36del) and p.(Arg64Cys), the episignature clustered both with the controls, which reinforces their classification as VUS and suggests that these individuals may have another disease or another mechanism; applied to the EpiSign knowledge base, it identified an undiagnosed individual with a score close to 1, who could not be recontacted [1]. The SET signature is balanced (hypo- and hypermethylation) and, in the similarity diagram between episignatures, lies on the same branch as Down and Williams syndromes [1].

Two practical limitations constrain its use. The first is that a negative episignature result does not exclude the pathogenicity of a variant acting through another mechanism, for example a missense variant with a partial effect, and that validation included only 7 training cases. The second is that the analysis was developed in the laboratory of Sadikovic (London Health Sciences Centre, Canada), who is a shareholder in EpiSign Inc.; EpiSign is commercial software, and the sources consulted do not state whether the SET signature is already part of the clinical panel available for request [1].

5.4. Test limitations and a negative result

A negative exome does not exclude the disease. Deletions of a few exons or of the whole gene may escape exome CNV analysis, which has lower sensitivity than targeted methods; this methodological limitation has been described for neighbouring STXBP1 and applies to SET, where deletions confined to the gene or to the critical region have been documented [4, 19]. Deep intronic variants are not detected by exome. Analysis and reporting on a non-MANE transcript, such as NM_001122821, may lead to overinterpretation of variants in exon 1 unique to isoform 1, and the reverse is also possible [4, 11]. Parental mosaicism was not documented in the sources read, but testing of parental leucocyte DNA does not detect mosaicism confined to the germline [17]. Finally, detection of a SET variant does not close the investigation if there are features outside the spectrum: the Zhang infant had an STXBP1 variant that explained the epilepsy, and multiple molecular diagnoses are described in neurodevelopmental disorders [21, 31]. Periodic reanalysis of non-diagnostic exome data allows identification of genes described later, as happened with SET [4].

5.5. Report content

The report should state the gene, the transcript with its version (preferably MANE NM_003011.4, with the NM_001122821 notation in brackets where useful for comparison with the literature), the HGVS c. and p. nomenclature, the genomic position with the genome build, the zygosity (heterozygous), the inheritance when known, the ACMG/AMP class with the criteria applied and, for CNVs, the coordinates, size, genes included and ACMG/ClinGen score [11, 13, 38]. For a VUS, the report should recommend parental testing and mention the possibility of functional study by episignature [1].

5.6. Practical algorithm

5.7. Prenatal and preimplantation diagnosis

Once the variant has been identified in the family, prenatal diagnosis by chorionic villus sampling or amniocentesis and preimplantation genetic diagnosis are possible [17]. There are no characteristic ultrasound findings: intrauterine growth restriction was rare (1 in 24) and unilateral renal agenesis was described once [1, 5]. The Shi series includes a 26-week fetus, with no published details on the indication for the study [1]. In the absence of a family history, the disease would be detected prenatally only by fetal exome or genome performed for another indication.

5.8. 9q34.11 duplications: detection, confirmation and classification

The 9q34.11 microduplication is detected by chromosomal microarray, whose resolution depends on probe density in the region, or by CNV analysis of the exome or genome, which on exome has lower sensitivity than targeted methods (section 5.4) [13, 19]. The microarray measures copy number but does not show where the additional copy lies. The report should state the cytoband, size, minimum and, when useful, maximum coordinates with the genome build, the copy state and the mechanism when known (for example, tandem duplication); determining the mechanism may require additional methods, such as metaphase FISH or long-read sequencing, already used to delineate deletions of the region [4, 13]. For terminal 9q34→qter duplications, parental testing must include karyotype or FISH, because a parent with a cryptic balanced translocation has a normal microarray [45, 46].

Parental testing is the second mandatory step. Many CNVs are polymorphisms without pathogenic significance, and several recurrent CNVs have incomplete penetrance and variable expressivity, so clinical assessment of the carrier parent is part of the interpretation [45, 47]. The table summarises the application of the ACMG/ClinGen standards for copy-number gains to a 9q34.11 microduplication including SET, with the curation status as of November 2026.

The result depends on the number of literature patients with confirmed de novo occurrence, which the sources consulted do not specify for the De Falco series. Counting only the index case, a small de novo duplication in a proband with ID scores 0.30 points (0.15 for the published case and 0.15 for de novo occurrence in the proband) and remains a VUS; six confirmed de novo published cases, scored under 4C, would be needed to reach the section 4 maximum of 0.90 and a total of 1.05 points (pathogenic). A duplication inherited from an asymptomatic parent moves down the VUS scale. Classification of the specificity of the phenotype is a laboratory judgement, and a review of SET triplosensitivity by ClinGen would change criterion 2A; the classification should therefore be revised when the curation changes [10, 13].

In prenatal diagnosis, some laboratories do not report CNVs with substantial incomplete penetrance, because of prognostic uncertainty and because they are unlikely to explain the ultrasound findings [45]. A 9q34.11 microduplication including SET, as a rule classified as a VUS, falls within this discussion: the decision to report it depends on the laboratory's policy and on the couple's prior consent, and counselling should explain that penetrance and severity have not been established.

6. Differential Diagnosis

6.1. General principle

When ID and speech impairment are not accompanied by distinctive findings, all autosomal dominant, autosomal recessive and X-linked forms of ID enter the differential diagnosis, and only genetic investigation separates them [17]. The differential diagnosis that is useful in practice is that of the entities that may be confused with SET-related disorder in the molecular result (neighbouring genes, dosage alterations of the same region, genes with similar names) or in the mechanism (other regulators of histone acetylation), and that of the syndromes clinically considered before diagnosis in the published patients.

6.2. Comparative table

6.3. Neighbouring genes and dosage alterations of the region

The first differential problem is the extent of the alteration. A deletion that removes SET and STXBP1 has a prognosis dominated by STXBP1 epilepsy, which starts at a median of 6 weeks of life and, in most, in the first year [19]. Inclusion of SPTAN1 raises the question of its contribution: pathogenic SPTAN1 variants cause early epileptic encephalopathy through a dominant-negative effect, and it is not known whether isolated SPTAN1 haploinsufficiency produces a severe phenotype [3]. The available data are contradictory: the 10 DECIPHER cases with SET and SPTAN1 had no seizures, but the individual in the Shi series with the 400 kb deletion including both genes had epilepsy [1, 5]. The deletion should therefore be delineated precisely and counselling about epilepsy should reflect the genes included. The 9q34.11 microduplication is the mirror image in copy gain: the De Falco patients have speech delay and ID like patients with SET loss, but with different dysmorphic features (midface hypoplasia, thin lips) [6]; the diagnosis is made by microarray and interpretation is discussed in section 2.8.

6.4. SETBP1: the same axis in opposite directions

SETBP1 disorders are the mechanistically closest differential. The SETBP1 protein binds SET and protects it from degradation. In Schinzel-Giedion syndrome, gain-of-function variants in a 12 bp degron of exon 4 prevent SETBP1 degradation and cause accumulation of SET, histone hypoacetylation and chromatin compaction at enhancers; the picture is much more severe than that of MRD58, with moderate to profound ID, refractory epilepsy, malformations and neuroepithelial neoplasms in about a quarter of patients [1, 18, 36]. SETBP1 haploinsufficiency (MRD29) would be expected to reduce SET stability, approaching MRD58, and does indeed show clinical overlap: mild dysmorphic features, global delay, ID and behavioural changes [1, 17]. The feature that most distinguishes MRD29 is severe childhood apraxia of speech, present in 80%, which can leave children with little verbal language for years [17]. Richardson and colleagues found no distinguishing features between patients with SET loss of function and those with SETBP1 [5], and GestaltMatcher did not cluster patients with SET with any of the three SETBP1 groups [1]. The distinction is exclusively molecular.

Schinzel-Giedion syndrome was delineated in 1978 from two siblings with severe midface retraction, skull base anomalies, heart disease, hydronephrosis and clubfeet; the sibling recurrence, initially difficult to interpret, is compatible with gonadal mosaicism, given that the SETBP1 variants identified by exome sequencing in 2010 were de novo in all 13 patients studied, all meeting the Lehman criteria [51]. These criteria require the facial phenotype (prominent forehead, midface retraction, short upturned nose), developmental delay beyond the neonatal period and hydronephrosis or two of four skeletal anomalies: sclerotic skull base, wide occipital synchondrosis, increased density or cortical thickness of the long bones and broad ribs [51]. Other supporting signs are alacrima with corneal hypoaesthesia, bilateral non-obstructive megacalycosis and stapes malformation; the course described includes infantile spasms with hypsarrhythmia, progressive neurological deterioration with involvement of grey and white matter, and embryonal tumours, mainly sacrococcygeal teratoma [51]. Santos and colleagues described a patient with the typical picture, hypothyroidism and diabetes insipidus, which they attributed to posterior fossa malformation and arachnoid cysts [51]. In MRD29, the association with SETBP1 loss of function resulted from a CNV morbidity map built from 29,085 children with developmental delay and 19,584 controls, complemented by resequencing of candidate genes; in the patients of Coe and colleagues, the cognitive spectrum ranged from normal intelligence with speech impairment to profound ID, and the most frequently reported dysmorphic pattern was an elongated face with distinctively shaped eyebrows [49]. None of these MRD29 features is specific enough to separate it clinically from MRD58.

6.5. Other disorders of histone acetylation

Shi and colleagues organised the mechanistic differentials into two groups: disorders that, like SET loss, would be expected to lead to hyperacetylation (MRD29 and the Menke-Hennekam syndromes due to gain of function of CREBBP and EP300) and those that lead to hypoacetylation (Schinzel-Giedion syndrome, the Rubinstein-Taybi syndromes due to loss of function of CREBBP and EP300 and Cornelia de Lange syndrome 1 due to NIPBL) [1]. Neither group shared facial similarity with the patients with SET, and similarity within the groups themselves was low, which prevents concluding that the direction of the acetylation change translates into a common facial phenotype [1]. Stevens adds that SET interacts with the proteins encoded by EP300, CREBBP, SETBP1, KMT2A, RAC1 and CTCF, all autosomal dominant ID genes [12].

Unlike MRD58, these syndromes have clinical signs that often allow targeted suspicion. Rubinstein-Taybi syndrome combines ID, postnatal growth deficiency, microcephaly, broad thumbs and halluces and a facies with arched eyebrows, long eyelashes, downslanting palpebral fissures, a beaked nose with the columella extending below the alae nasi and a grimacing smile, together with a propensity to keloids and an increased tumour risk [52]. The CREBBP form (RSTS1) accounts for 50 to 70% of cases and the EP300 form (RSTS2) for about 3%; RSTS2 tends to have milder cognitive impairment, less marked facial dysmorphism and more pronounced microcephaly, and broad thumbs and halluces may be absent, so their absence does not exclude the diagnosis [53]. The birth prevalence cited is 1 in 100,000 to 125,000; CREBBP deletions, from a few hundred base pairs to several megabases and sometimes in low-level mosaicism, justify a dosage study when sequencing is negative, and the entry also records an old anaesthetic warning about cardiac arrhythmia with suxamethonium [52]. The Menke-Hennekam syndromes result from variants in a restricted segment of CREBBP (MKHK1) or EP300 (MKHK2), in the final part of exon 30 and the initial part of exon 31, de novo in all patients with parents studied; patients have ID of variable degree, autistic behaviour in more than half, microcephaly, short stature, feeding difficulties, hearing impairment and recurrent upper respiratory infections, without the broad thumbs or the facies of Rubinstein-Taybi syndrome [54, 55]. MKHK2 has been described in few patients, and one of the two girls described by Menke and colleagues had a full-scale IQ of 91, which shows that ID is not constant [55]. In Cornelia de Lange syndrome 1, the facies (low frontal hairline, arched eyebrows with synophrys, anteverted nares, long philtrum, thin lips with downturned corners), prenatal and postnatal growth restriction and upper limb reduction defects guide the diagnosis in the classic forms; NIPBL accounts for 50 to 60% of cases, and in a survey of 65 dysmorphologists the classic forms were recognised in 90% and the mild forms in 54% [56]. In a Dutch cohort, 10 of 44 patients (23%) had the NIPBL variant detectable in buccal mucosa cells and not in lymphocytes, which supports repeating the study in another tissue when clinical suspicion persists [56]. In a review of 426 deaths occurring between 1966 and 2007, respiratory causes, including aspiration and reflux, were the most frequent (31% of the 295 with a known cause), followed by gastrointestinal causes (19%) [56]. In a patient with mild ID, without major malformations and without a recognisable facies, none of these pictures is likely, and the diagnosis of MRD58 continues to depend on sequencing.

6.6. Clinical differentials considered before diagnosis

Williams syndrome and Pitt-Hopkins syndrome were considered in the DDD patients before the exome [5], and atypical Rett syndrome was the working diagnosis of the Condell patient during childhood, with negative studies of MECP2, CDKL5 and FOXG1; the patient meets several Rett supportive criteria (bruxism, small cold hands and feet, reduced response to pain) [4, 43]. The Hamdan case, with microcephaly, led the authors to propose a resemblance to MCPH1 primary microcephaly, based on the SET–MCPH1 interaction in chromosome condensation [5, 28]. The name of the gene also creates bibliographic confusion with genes with a SET domain, which encode histone methyltransferases and are unrelated to the SET gene; a 2026 review of these genes deals with another protein family [5, 21].

7. Treatment and Risk Management Strategies

7.1. Framework: absence of specific therapy and of a dedicated guideline

There is no treatment targeted at SET loss of function and no published management guideline for MRD58; a ClinicalTrials.gov search in October 2026 found no dedicated studies, and the literature search found no specific preclinical therapies, which does not prove that unpublished projects do not exist [1, 21]. Management is supportive and rests on two pillars. The first is the observations from the SET series, especially the value of early and intensive speech therapy and the frequency of ocular, sleep and behavioural problems [1, 4]. The second is the declared extrapolation of the GeneReviews recommendations for SETBP1 haploinsufficiency disorder, a condition with an overlapping phenotype (ID, severe speech impairment, behaviour, ophthalmological problems) and a close mechanism; the authors of that GeneReviews chapter themselves state that their recommendations are based on the literature and clinical experience, without clinical practice guidelines [17]. The drug doses given below come from the official labels for the corresponding indications (insomnia in ASD and in neurogenetic diseases, irritability associated with autism, ADHD, focal epilepsy), because there are no published doses for MRD58; none of these medicines has been studied specifically in this disease.

7.2. Initial assessment after diagnosis

The proposed initial assessment follows the model of the GeneReviews assessment table for SETBP1 haploinsufficiency, adapted to the findings described in MRD58 [1, 17].

In patients with a 9q34.11 microduplication, the sources do not describe specific malformations or organ complications, and there is no specific treatment. Initial assessment follows the domains in the table, with emphasis on development, speech, behaviour and vision, and symptomatic treatment of comorbidities uses the same indications and doses as in section 7.4, by declared extrapolation from the label of each medicine for the condition treated (insomnia, ADHD, irritability in ASD or epilepsy), and not from microduplication data [6, 17]. The genetic assessment adds parental testing, definition of the mechanism of the duplication when it has reproductive consequences and a commitment to reanalysis of the classification (section 5.8) [13].

7.3. Non-pharmacological interventions

Speech therapy is the intervention with the most support in the SET sources. Condell and colleagues attribute to intensive and early speech therapy the recovery of phrase-level communication in adulthood in their patient, who had no functional speech in childhood, and recommend early speech assessment and intervention sustained over time, acknowledging that the most effective type of therapy has not been studied [4]. By extrapolation from MRD29, therapy adjusted to the developmental profile and age is recommended, with early phonological awareness tasks to support reading and writing and, where there is apraxia of speech, intensive evidence-based motor speech therapies [17]. Augmentative and alternative communication, from picture exchange to speech-generating devices, should be introduced early: according to the GeneReviews chapter on MRD29, it does not delay verbal speech and supports speech and language development [17].

Early intervention should include physiotherapy and occupational therapy for hypotonia and motor delay and, when necessary, feeding therapy with texture modification and a low threshold for videofluoroscopic study if aspiration is suspected [17]. In the first 3 years, early intervention programmes bring together occupational therapy, physiotherapy, speech and feeding therapy and special education; from 3 to 5 years, the GeneReviews model provides for preschool education with an individual plan, which in Portugal and Brazil corresponds to the national early intervention and inclusive education systems [17]. For behaviour, standardised interventions apply, including the behavioural therapies used in ASD where relevant, and an educational plan that addresses behaviour [17]. Refractive errors, strabismus and amblyopia are treated conventionally [1, 17]. In adolescence, there should be a plan for transition to adult care, vocational training and supported housing, as in the Condell case [4, 17].

7.4. Pharmacological treatment of comorbidities

The medicines below are used for a symptom or associated condition and not for an approved indication for MRD58; their use in this disease is therefore an extrapolation, and the decision follows the standards of the specialty treating the comorbidity. In the absence of a specific guideline, the reference doses are those of the official labels consulted, all from the United States except that of prolonged-release melatonin and the harmonised SmPC for risperidone, which are European; Portuguese summaries of product characteristics (SmPC) and Brazilian package inserts were not consulted for this edition.

Prolonged-release melatonin (Slenyto®) for insomnia. Melatonin acts on MT1, MT2 and MT3 receptors, mainly the first two, which are involved in the regulation of circadian rhythms and sleep [57]. In the European Union, Slenyto® is indicated for insomnia in children and adolescents aged 2 to 18 years with ASD and/or neurogenetic disorders with aberrant daytime melatonin secretion and/or nocturnal awakenings, when sleep hygiene measures have been insufficient, and for insomnia in patients aged 6 to 17 years with ADHD [57]. The starting dose is 2 mg, increased to 5 mg if the response is inadequate, with a maximum of 10 mg, once daily, 30 to 60 minutes before bedtime, with or after food; the tablets are swallowed whole, without breaking, crushing or chewing, and may be mixed with yoghurt, orange juice or ice cream and taken immediately [57]. The effect should be assessed after at least 3 months, considering discontinuation if there is no clinically relevant benefit, and treatment should be reviewed at least every 6 months [57]. The most frequent adverse effects (between 1 in 100 and 1 in 10 children) were somnolence, fatigue, mood changes, headache, irritability, aggression and morning hangover; it is not recommended in hepatic impairment and is used with caution in renal impairment; beta blockers may suppress nocturnal endogenous melatonin secretion and should be taken in the morning [57]. Sleep disorder affects 41% of patients with MRD58, but the medicine has not been studied in this disease; the label indication of «neurogenetic disorders» is generic and MRD58 was not included in the trials [1, 57].

Risperidone (Risperdal®) for severe irritability. Risperidone is an atypical antipsychotic whose therapeutic effect is attributed to antagonism of dopamine D2 and serotonin 5-HT2 receptors, with a contribution from the metabolite 9-hydroxyrisperidone (paliperidone) [58]. In the United States, it is approved for irritability associated with autism, including aggression, deliberate self-injury, temper tantrums and quickly changing moods, with efficacy demonstrated in three short-term trials in children and adolescents aged 5 to 17 years [58]. The starting dose is 0.25 mg/day under 20 kg and 0.5 mg/day from 20 kg; after at least 4 days, it is increased to 0.5 mg/day and 1 mg/day, respectively, and maintained for at least 14 days; if the response is insufficient, it is increased at intervals of 2 weeks or more, by 0.25 mg/day under 20 kg and 0.5 mg/day from 20 kg; the effective range is 0.5–3 mg/day, with no maximum dose defined in the source beyond this range, as a single dose or divided into two, and there are no dosing data below 15 kg [58]. In paediatric trials of 3 to 8 weeks, mean weight gain was 2 kg versus 0.6 kg with placebo, with a gain of at least 7% in 33% versus 7%, and of 5.5 kg at 24 weeks and 8 kg at 48 weeks in the open-label studies; prolactin rose in 49% versus 2%; somnolence was frequent, especially at the start; tardive dyskinesia occurred in 0.1% of 1885 children and adolescents and reversed on discontinuation [58]. In the European Union, risperidone has no indication in ASD: it is authorised from 5 years only for short-term (up to 6 weeks) symptomatic treatment of persistent aggression in conduct disorder, in children and adolescents with below-average intellectual functioning or intellectual disability, with a starting dose of 0.25 mg/day under 50 kg (optimum 0.5 mg/day; maximum 0.75 mg/day) and 0.5 mg/day from 50 kg (optimum 1 mg/day; maximum 1.5 mg/day) [59]. Weight should be monitored and the benefit-risk balance reassessed periodically in long-term use [58]. In MRD58, risperidone was used in the adult Condell patient, in combination, for anxiety, with a good result according to the authors; this is an isolated report and an indication different from that on the label [4].

Methylphenidate (Ritalin®) for ADHD. Methylphenidate is a central nervous system stimulant; the US label states that its therapeutic mode of action in ADHD is not known [60]. The US label for the immediate-release tablet indicates it for ADHD from 6 years and in adults [60]. In children from 6 years, it is started at 5 mg twice daily, before breakfast and lunch, with weekly increases of 5 to 10 mg; doses above 60 mg/day are not recommended; in adults, the average dose is 20–30 mg/day in 2 or 3 doses, with a maximum of 60 mg/day [60]. Before starting, cardiac disease is assessed (personal history, family history of sudden death or ventricular arrhythmia, physical examination) as well as the presence of tics or Tourette syndrome; if there is paradoxical worsening or adverse effects, the dose is reduced or discontinued, and if there is no improvement after one month of adjustment, it is discontinued [60]. The label carries a boxed warning on abuse, misuse and addiction, which requires assessment of risk, education of the family on storage and disposal and monitoring for signs of misuse [60]. ADHD or hyperactivity was described in 6 of 21 individuals in the Shi series and in patients of Richardson and Nishi [1, 2, 5].

Levetiracetam (Keppra®) for focal seizures. The antiseizure mechanism of levetiracetam has not been elucidated; the drug binds the synaptic vesicle protein SV2A, involved in exocytosis, and affinity for this protein correlates with anticonvulsant potency in animal models [61]. It was the medicine that controlled the focal seizures of Mai patient 1, with no dose reported [21]. The choice of antiseizure medicine in MRD58 follows the seizure type, as in epilepsy generally; no medicine has been studied specifically in this disease, a principle that the GeneReviews chapter on MRD29 also states for that disorder [17]. In the United States, levetiracetam is indicated for focal seizures from 1 month of age [61]. For children from 6 months to less than 4 years, the starting dose is 10 mg/kg twice daily, increasing by 10 mg/kg twice daily after 2 weeks up to 25 mg/kg twice daily; from 4 to 16 years, it is started at 10 mg/kg twice daily and increased every 2 weeks up to 30 mg/kg twice daily, with a maximum of 3000 mg/day; from 16 years, 500 mg twice daily, with increases of 500 mg twice daily every 2 weeks up to 1500 mg twice daily; the oral solution is preferred under 20 kg, and the dose is adjusted to renal function [61]. The label highlights behavioural effects: non-psychotic behavioural symptoms in 38% of treated children aged 4 to 16 years versus 19% with placebo, irritability in 12% versus 0% between 1 month and 4 years, and psychotic symptoms in 17% versus 5% in this age group; discontinuation should be gradual [61]. Given that almost half of patients with MRD58 have behavioural changes, this profile should be explained to the family and monitored [1, 61].

Other medicines mentioned in reports. In the adult Condell patient, anxiety was treated with pregabalin, risperidone and clobazam, and constipation with various laxatives, with no published doses [4]. These are individual management choices that do not constitute recommended practice for the disease.

7.5. Comparative regulatory status

No medicine has an approved indication for MRD58, in the United States or in the European Union. The table summarises the status of the symptomatic medicines used as reference.

7.6. Reference clinical trials and their limitations

There are no trials in MRD58. The trials supporting the labels used as reference were conducted in other populations.

The results of these trials cannot be transposed to MRD58 without reservations. The populations did not include patients with MRD58; the Slenyto® SmPC does not give confidence intervals for the differences, only p values; the absolute difference of 33 minutes in total sleep should be read in the light of baseline sleep duration, not described in the extract; there is no direct comparison between medicines, and any indirect comparison between trials in different populations is not valid. The sources consulted did not identify negative trials, suspensions or withdrawals relevant to these indications, but the search was not directed to that end.

7.7. Developments in the last 12 months and research

All the developments are diagnostic and natural history in nature. No regulatory applications, human trials or preclinical therapies targeted at SET are described in the sources consulted. The Zaghi study on SET accumulation in Schinzel-Giedion syndrome is preclinical research on the reverse mechanism and does not propose a treatment for MRD58 [36].

7.8. Agents to avoid, pregnancy, lactation, perioperative care and transition

There are no medicines contraindicated because of the disease. The precautions derive from the comorbidities: levetiracetam may worsen behaviour, which is already abnormal in almost half of patients, and risperidone causes weight gain and hyperprolactinaemia [1, 58, 61]. Pregnancy is documented in one affected woman who had an affected son, with no data on obstetric complications [4, 12]. For melatonin, the SmPC recommends, as a precaution, avoiding use in pregnancy, because of the absence of data in pregnant women, and deciding between stopping breastfeeding or the medicine, because melatonin passes into milk in animal studies [57]. The sources describe no anaesthetic or perioperative complications; hypotonia, recurrent respiratory infections (23%) and feeding problems justify the usual pre-anaesthetic assessment of the child with ID [1]. Transition to adult care should provide for continuity of speech therapy, which may bring gains into adulthood, monitoring of constipation, the spine and mental health, and planning of autonomy and housing [4, 17].

7.9. Available recommendations and divergences

There are no scientific society guidelines on MRD58, so there are no divergences between societies to record. The applicable recommendations are the general ones: the ACMG guideline for the use of exome and genome sequencing in ID [23], the standards for interpretation of sequence variants and CNVs [13, 38] and the management recommendations for MRD29, used by extrapolation [17]. The only relevant divergence between sources concerns sensitivity to SET triplication, discussed in section 2.8 [6, 10].

8. Reproductive Options and Family Counselling

8.1. Recurrence risks by scenario

MRD58 is autosomal dominant and arises de novo in the great majority of cases: 21 of 22 individuals with studied inheritance in the Shi series, and all the Richardson, Nishi and Mai patients with tested parents [1, 2, 5, 21]. In the Shi series there was no recurrence in siblings [1]. The counselling scenarios, constructed by extrapolation from the GeneReviews model for MRD29, which has the same genetic architecture, are as follows [17].

In the 9q34.11 microduplication, counselling starts from a double uncertainty: the causal relationship with SET triplication has not yet been accepted by ClinGen, and penetrance cannot be estimated from the published data [6, 10]. When the duplication is inherited from an asymptomatic parent, the transmission risk is 50%, but the probability that a carrier child will have a neurodevelopmental disorder is unknown; the population duplications in gnomAD indicate that there are carriers without severe disease [22]. For CNVs with incomplete penetrance, there is no consensus on cascade testing or prenatal diagnosis, and counselling should present these limitations before any reproductive decision [45]. In terminal duplications resulting from a parental balanced translocation, testing extends to the relatives of the carrier parent, because of the risk of imbalance in other pregnancies in the family [45, 46].

8.2. Testing of relatives and carriers

Testing of both parents is the first step after diagnosis: it confirms de novo occurrence, contributes to variant classification (criterion PS2) and defines the recurrence risk [17, 38]. When a parent has the variant, that parent and his or her siblings should be assessed clinically, because expressivity varies between borderline intellectual functioning and moderate ID, as in the mother described by Stevens, with moderate ID and a reproductive life [1, 4, 12]. Little clinical information was available on the probably affected father in the Shi series [1]. No carriers without manifestations have been described.

8.3. Prenatal and preimplantation diagnosis

Once the variant has been identified, prenatal diagnosis by chorionic villus sampling or amniocentesis and preimplantation genetic diagnosis can be offered both to the parents of a de novo case (because of the residual risk of germline mosaicism) and to an affected adult [17]. The decision rests with the couple, after counselling that explains the usually mild to moderate severity, the variability and the absence of characteristic ultrasound findings [1].

8.4. Counselling of affected adults and minors

Adults with mild ID, or with borderline functioning, should receive adapted information on the 50% risk to offspring and on reproductive options, ideally at the transition to adulthood [1, 17]. Testing clinically unaffected minor siblings has no medical utility, because the disease manifests early with developmental delay and there are no preventive measures that depend on genotype; in children with delay, testing is part of the diagnostic investigation. Counselling should also address the variability of prognosis between siblings or between parent and child with the same variant [1, 12].

9. Special Considerations: Population Diversity and Epidemiology

9.1. Frequency: what is known and what each number measures

There is no published estimate of the prevalence or incidence of MRD58. The only frequency figure in a defined cohort comes from the DDD study: 3 patients with SET variants in 4323 families with completed analysis, or 0.069% of a UK and Irish cohort referred for developmental disorders, which does not represent the general population [5]. In the Hamdan series, the variant was found in 1 of 41 Quebec trios with moderate or severe ID, a number too small to estimate frequency [28]. The context is that of ID in general, which affects about 1–2% of the population [4], and of neurodevelopmental disorders, which affect up to 3% [1]. Adding up the series without duplications, about 42 individuals with pathogenic or likely pathogenic sequence variants had been described by October 2026 (38 gathered by Shi, plus the Zhang, Nishi and Mai patients) and at least four with deletions of the region described in detail; this count is a sum of publications, not a prevalence, and omits cases recorded only in databases [1–4, 21, 31]. The apparent rarity also reflects the fact that diagnosis depends on exome or genome sequencing.

For the 9q34.11 microduplication there is also no published prevalence. The 14 patients described were gathered at centres in Italy, France, Poland and Australia, which reflects the origin of the consortium and not a population distribution [6]. In gnomAD v4, which includes genomes of several ancestries, three copies of duplications encompassing SET were observed among about 126,000 alleles, a population carrier frequency, not a disease prevalence [22]. The sources consulted contain no data from Portugal or Brazil for any of the entities in this article.

9.2. Geographical distribution and ancestry

The published patients come from Canada (Hamdan; the Stevens series, with the CAUSES study in Vancouver), the Netherlands and the United States (Stevens), the United Kingdom (DDD), Greece (Marinakis), Germany (Brunet), China (Zhang, Pan, Mai), Japan (Shono, Nishi), Australia (Condell) and a European and North American network (Shi: the Netherlands, United States, Spain, Switzerland, France, United Kingdom and Canada) [1, 2, 4, 5, 12, 21, 28, 32, 33]. The distribution follows access to exome sequencing, not a difference in frequency between populations. Because the disease results mainly from de novo variants, no founder effect is expected, and none has been described; the variant p.(Arg44LeufsTer10) has been described de novo in patients from the United Kingdom, Japan and several centres of the Shi series; it lies in a repeated sequence, which is compatible with recurrent mutation rather than a common origin, although the origin of the haplotypes has not been studied [1, 2, 11]. The sources consulted describe no patients from Portugal or Brazil, which does not demonstrate that none exist.

9.3. Special populations

Information on adults is scarce (three adults described), which limits counselling on ageing [4, 12]. There are no data on pregnancy in affected women beyond the maternal transmission described, nor on differences between sexes: the Shi series has 10 females and 13 males, with no differences reported [1].

9.4. European reference networks and resources for families

The relevant European network is ERN ITHACA (congenital anomalies and intellectual disability), which runs a cohort project on the SET gene coordinated in the Department of Genetics of CHU de Rouen, with the aim of delineating the neurological and developmental phenotype, characterising the dysmorphic features and assessing the course into adulthood [24]. Radboudumc maintains, in its Human Disease Genes series, a page on SET for professionals and families, which collects patient data and describes the disorder as multisystemic, with a very limited number of patients [25]. No patient organisation dedicated to SET was found; the apraxia of speech and general ID organisations referred to in the GeneReviews chapter on MRD29 serve as an alternative resource [17]. GeneMatcher made it possible to gather the largest series, which shows the value of case sharing for diseases of this rarity [1].

10. Illustrative Clinical Case

10.1. Main case: recurrent variant in a Japanese child

Nishi and colleagues described in 2026 a Japanese boy seen at 7 years [2]. He was conceived by in vitro fertilisation and born at 38 weeks and 1 day, with a weight of 3622 g (+1.5 standard deviations), length of 50.0 cm (+0.3) and head circumference of 35.0 cm (+1.2). He achieved head control at 5 months, independent walking at 24 months and first words at 22 months, with mild generalised hypotonia. On the Kyoto scale, at 7 years and 3 months, the overall developmental quotient was 66, with 43 in the postural-motor domain, 68 in the cognitive-adaptive domain and 67 in the language-social domain, corresponding to mild ID with more marked motor delay. At 7 years and 4 months, height was 118.6 cm (−0.6), weight 24.8 kg (+0.5) and head circumference 52.5 cm (+0.3). Examination showed a prominent forehead, hypertelorism, downslanting palpebral fissures, periorbital fullness, synophrys, broad nasal bridge and tip, wide mouth, thick lower lip, widely spaced teeth with diastema, incomplete lip closure and tongue protrusion. He had mild hyperactivity and a cheerful, sociable temperament. Brain MRI and abdominal ultrasound were normal, and he had no seizures. The exome identified the variant NM_003011.4:c.128_131del p.(Arg44LeufsTer10), de novo, classified as pathogenic (PVS1, PS2, PS3, PM2) and already recorded in ClinVar (VCV000521454) [2, 11]. Figure 1 of the article shows the face and dentition of the two individuals described: https://pmc.ncbi.nlm.nih.gov/articles/PMC13422647/figure/cga70063-fig-0001/ [2].

The case brings together the typical features of the disease: global delay with hypotonia, first words and walking around 2 years, mild ID, hyperactivity, mild dysmorphic features and normal investigations [1]. The variant is the most frequent in the literature, described in the Stevens mother and son, in two DDD patients (as c.167_170del on NM_001122821) and in ten individuals of the Shi series [1, 5, 12]. The management described — developmental surveillance and educational support — is supportive, consistent with the approach in sections 7 and 11.

10.2. Contrasting case: minimal deletion and a 28-year diagnostic odyssey

Condell and colleagues described an Australian woman whose investigation lasted 28 years [4]. In infancy, she sat and rolled late (about 4.5 months), had global delay evident at 7 months, gross and fine motor regression that prompted physiotherapy and initial absence of functional speech. She progressed to severe ID, ASD, laughing and screaming spells, hand stereotypies, gait deterioration, sleep disturbance, bruxism, peripheral vasomotor changes with small cold hands and feet, episodic hyperventilation and reduced response to pain, a picture that led to a working diagnosis of atypical Rett syndrome. Studies of MECP2, CDKL5 and FOXG1 were negative, as were the initial trio exome, genome and microarray. MRI was normal and she had no seizures. The face showed asymmetry, strabismus from 2 years, a broad nasal tip and columella, synophrys, thick medial eyebrows and a chin dimple; head circumference was at the 75th centile [4].

Reanalysis identified a de novo deletion of about 326 kb at 9q34.11 (GRCh38 chr9:128656218-128981804), with 13 genes, confirmed by SNP microarray and nanopore long-read sequencing, without STXBP1 or SPTAN1, classified as pathogenic with 1.15 points under the ACMG/ClinGen standards [4, 13]. Fibroblast proteomics showed SET protein reduced to 50% [4]. At 28 years, severe ID had been stable for several years; she walked but tired (usually less than 200 m at a time); hand function had declined; speech was very limited, but with short phrases and sounds that the family interprets, after intensive speech therapy; she had significant constipation, kyphosis, hypotonia and two episodes of unexplained loss of consciousness with a normal EEG; anxiety was treated with pregabalin, risperidone and clobazam, with a good result according to the authors; she lived in supported residential care [4]. The article, with photographs from childhood to adulthood, the diagram of the interval and the proteomics, is available at https://pmc.ncbi.nlm.nih.gov/articles/PMC13533875/ [4].

The case does not represent the typical phenotype. Severe ID and Rett-like features have not been described in patients with point variants, and the authors consider that a contribution from other genes in the deletion has not been excluded. The drug combination used for anxiety is an individual decision and does not constitute a recommendation. The case illustrates, however, three useful points: reanalysis of genomic data can resolve long-standing diagnoses, speech can progress in adulthood and a small deletion can escape initial exome and microarray analyses [4].

11. Prognosis and Long-Term Follow-up

11.1. Prognosis

Life expectancy has not been studied, and the sources describe no deaths attributable to the disease or severe malformations that would limit it [1]. The functional prognosis is, for most patients with point variants, that of mild ID or borderline functioning: in the Shi series, 94% of individuals speak and 88% walk, and those who did not yet do so were under 30 months [1]. Speech remains the most persistent deficit, but it progresses, and there are reports of gains into adulthood [4]. Deletions with severe ID have a worse functional prognosis [4]. The schooling of the DDD patients took place in special school, in one case combined with mainstream education [5]. Independence in adulthood was partial in the few adults described: the mother of the Stevens family had a reproductive life and reads simple texts, and the Condell patient lives with support [4, 12]. The comorbidities that weigh most in follow-up are behavioural changes (48%), sleep disorder (41%), visual problems (74%) and seizures (17.9%) [1].

11.2. Surveillance

The surveillance recommendations follow the GeneReviews model for MRD29, in which intervals are set by the responsible specialist and not by a fixed schedule, adapted to the findings in MRD58 [1, 17].

The sources do not justify systematic cancer, cardiac or renal screening: the SET fusions of leukaemia are somatic, and cardiac and renal malformations were isolated findings [1, 5, 34].

12. Summary and Concluding Remarks

SET-related intellectual developmental disorder (MRD58) results from haploinsufficiency of SET, a 9q34.11 gene encoding a component of the inhibitor of histone acetyltransferases complex (INHAT) and a histone chaperone. The disease almost always arises de novo, through truncating, splicing, rare missense variants or deletions including the gene, and was delineated between 2014 and 2018; about 42 patients with sequence variants and a few deletions had been published by October 2026, almost half in the 2026 Shi series.

The clinical picture is one of global developmental delay with axial hypotonia, first words around 2 years, walking between 2 and 3 years and progression to mild ID or borderline functioning, with ocular problems, sleep disorder, behavioural changes and, in a minority, seizures. The face is not recognisable, even by computational analysis, and there are no biochemical or imaging markers; diagnosis depends on exome or genome sequencing with CNV analysis, and the DNA methylation episignature described in 2026 allows VUS to be assessed. Interpretation requires attention to the transcript: the older literature uses NM_001122821, MANE is NM_003011.4, and variants in exon 1 unique to isoform 1 have no MANE equivalent and should not be treated as loss of function without proof. Deletions that include STXBP1 or SPTAN1 carry their own epilepsy risk, and severe early epilepsy in a child with a SET variant should prompt a search for another cause. 9q34.11 microduplications encompassing SET were associated in 2026 with a neurodevelopmental disorder with speech delay, ID and recurrent dysmorphic features in 14 patients, but SET triplosensitivity still has a score of 0 in ClinGen and duplications confined to SET exist in individuals in population databases; these duplications are classified under the CNV standards for gains, as a rule as VUS, with parental testing and scheduled reanalysis, and are distinguished from terminal 9q34→qter duplications and from 9q34.3 microduplications including EHMT1.

There is no specific treatment. Management consists of rehabilitation, with early and prolonged speech therapy, augmentative communication, physiotherapy and educational support, and symptomatic treatment of sleep, behaviour and seizures according to the label indications and doses for those conditions, an extrapolation that should be explained to families. The recurrence risk for the parents of a de novo case is low, and that of an affected adult transmitting the variant is 50%. The ERN ITHACA cohort and future series should clarify the course in adulthood, the genotype–phenotype correlation and the clinical utility of the episignature.

13. References

1. Shi Y, Silva A, Debuy C, Ghosh S, McConkey H, Schot R, et al. Delineating the clinical and molecular spectrum of the neurodevelopmental disorder associated with SET. Genet Med. 2026;28(9):102637. doi:10.1016/j.gim.2026.102637. PMID: 42322191.
2. Nishi E, Yanagi K, Kaname T, Okamoto N. Clinical and genetic findings in two Japanese individuals with SET-related neurodevelopmental disorder. Congenit Anom (Kyoto). 2026;66(1):e70063. doi:10.1002/cga.70063. PMID: 42141719; PMCID: PMC13422647.
3. Shono K, Enomoto Y, Tsurusaki Y, Kumaki T, Masuno M, Kurosawa K. Further delineation of SET-related intellectual disability syndrome. Am J Med Genet A. 2022;188(5):1595-9. doi:10.1002/ajmg.a.62681. PMID: 35122673.
4. Condell A, Zhang E, Sikora T, Massey S, Van Bergen NJ, Wang M, et al. Elucidating the role of SET as a key contributor to neurodevelopmental disability within the 9q34.11 deletion syndrome interval. Clin Genet. 2026;110(4):438-48. doi:10.1111/cge.70213. PMID: 42432431; PMCID: PMC13533875.
5. Richardson R, Splitt M, Newbury-Ecob R, Hulbert A, Kennedy J, Weber A; DDD Study. SET de novo frameshift variants associated with developmental delay and intellectual disabilities. Eur J Hum Genet. 2018;26(9):1306-11. doi:10.1038/s41431-018-0199-y. PMID: 29907757; PMCID: PMC6117329.
6. De Falco A, Vincent M, Vieville G, Gauthier M, Dieterich K, Coutton C, et al. 9q34.11 microduplications encompassing SET gene are associated with neurodevelopmental disorder and recurrent dysmorphisms. Am J Med Genet A. 2026;200(3):706-17. doi:10.1002/ajmg.a.64303. PMID: 41267543.
7. Orphanet. Distal duplication 9q syndrome (distal trisomy 9q). ORPHA:96101 [Internet]. Disease page. Accessed 1 Nov 2026.
8. Rots D, Rooney K, Relator R, Kerkhof J, McConkey H, Pfundt R, et al. Refining the 9q34.3 microduplication syndrome reveals mild neurodevelopmental features associated with a distinct global DNA methylation profile. Clin Genet. 2024;105(6):655-60. doi:10.1111/cge.14498. PMID: 38384171.
9. OMIM® Online Mendelian Inheritance in Man. *600960 SET NUCLEAR PROTOONCOGENE; SET. Johns Hopkins University; last edited 29/04/2022. https://omim.org/entry/600960. Accessed 28 October 2026.
10. ClinGen. SET (HGNC:10760) — gene-disease validity (intellectual disability, Definitive, 17/11/2021) and dosage sensitivity (HI 3; TS 0). https://search.clinicalgenome.org/kb/genes/HGNC:10760 (accessed 2026-10-28).
11. Freeman PJ, Hart RK, Gretton LJ, Brookes AJ, Dalgleish R. VariantValidator: accurate validation, mapping, and formatting of sequence variation descriptions. Hum Mutat. 2018;39(1):61-8. doi:10.1002/humu.23348. PMID: 28967166. VariantValidator 4.0.1 (vvdb_2025_3), GRCh38, accessed 2026-10-28 and 2026-11-01.
12. Stevens SJC, van der Schoot V, Leduc MS, Rinne T, Lalani SR, Weiss MM, et al. De novo mutations in the SET nuclear proto-oncogene, encoding a component of the inhibitor of histone acetyltransferases (INHAT) complex in patients with nonsyndromic intellectual disability. Hum Mutat. 2018;39(7):1014-23. doi:10.1002/humu.23541. PMID: 29688601.
13. Riggs ER, Andersen EF, Cherry AM, Kantarci S, Kearney H, Patel A, et al. Technical standards for the interpretation and reporting of constitutional copy-number variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics (ACMG) and the Clinical Genome Resource (ClinGen). Genet Med. 2020;22(2):245-57. doi:10.1038/s41436-019-0686-8. PMID: 31690835; PMCID: PMC7313390.
14. OMIM® Online Mendelian Inheritance in Man. #618106 INTELLECTUAL DEVELOPMENTAL DISORDER, AUTOSOMAL DOMINANT 58; MRD58. Johns Hopkins University; last edited 29/04/2022. https://omim.org/entry/618106. Accessed 28 October 2026.
15. Orphanet. Autosomal dominant non-syndromic intellectual disability (ORPHA:178469). https://www.orpha.net/en/disease/detail/178469
16. Meza-Espinoza JP, González-García JR, Patrón-Baro LI, González-Arreola RM, Contreras-Gutiérrez JA, Camberos-Barraza J, et al. Aphallia in a patient with 9q34 duplication syndrome: a case report. BMC Urol. 2025;25(1):133. doi:10.1186/s12894-025-01818-3. PMID: 40399931; PMCID: PMC12096471.
17. Morgan A, Srivastava S, Duis J, van Bon B. SETBP1 Haploinsufficiency Disorder. 2021 Nov 18 [revised 2024 May 9]. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–. Available from: https://www.ncbi.nlm.nih.gov/books/NBK575336/
18. Duis J, van Bon BWM. Schinzel-Giedion Syndrome. 2024 Mar 7. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–. Available from: https://www.ncbi.nlm.nih.gov/books/NBK601394/
19. Mercimek-Andrews S. STXBP1 Encephalopathy with Epilepsy. 2016 Dec 1 [updated 2023 Sep 28]. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–. Available from: https://www.ncbi.nlm.nih.gov/books/NBK396561/
20. ClinGen. Dosage Sensitivity Curations: gene curation list and region curation list, GRCh38 [Internet]. Clinical Genome Resource; accessed 2026-11-01. Available from: https://ftp.clinicalgenome.org/ClinGen_gene_curation_list_GRCh38.tsv and https://ftp.clinicalgenome.org/ClinGen_region_curation_list_GRCh38.tsv
21. Mai J, Ao L, Zhang Q, Liu L, Zeng Q, Luo X, Cao D, Chen L, Wen F. Phenotypic delineation of SET-related neurodevelopmental disorder: two case reports and literature review. Rare Dis Orphan Drugs J. 2026;5:13. doi:10.20517/rdodj.2025.89.
22. gnomAD. SET (ENSG00000119335): structural variants, gnomAD SV v4, GRCh38 [Internet]. Cambridge (MA): Broad Institute; accessed via the GraphQL API on 2026-11-01. Available from: https://gnomad.broadinstitute.org/gene/ENSG00000119335?dataset=gnomad_sv_r4
23. Manickam K, McClain MR, Demmer LA, Biswas S, Kearney HM, Malinowski J, et al. Exome and genome sequencing for pediatric patients with congenital anomalies or intellectual disability: an evidence-based clinical guideline of the American College of Medical Genetics and Genomics (ACMG). Genet Med. 2021;23(11):2029-37. doi:10.1038/s41436-021-01242-6. PMID: 34211152.
24. ERN ITHACA. SET gene: characterization of the clinical phenotype of patients carrying pathogenic variations of the SET gene (cohort project; coord. C. Saurin, CHU Rouen). https://ern-ithaca.eu/?p=8741 (accessed 2026-10-28).
25. Radboudumc. Human Disease Genes website series — SET. https://humandiseasegenes.nl/set/ (accessed 2026-10-28).
26. von Lindern M, van Baal S, Wiegant J, Raap A, Hagemeijer A, Grosveld G. Can, a putative oncogene associated with myeloid leukemogenesis, may be activated by fusion of its 3' half to different genes: characterization of the set gene. Mol Cell Biol. 1992;12(8):3346-55. doi:10.1128/mcb.12.8.3346-3355.1992. PMID: 1630450; PMCID: PMC364582.
27. Seo SB, McNamara P, Heo S, Turner A, Lane WS, Chakravarti D. Regulation of histone acetylation and transcription by INHAT, a human cellular complex containing the set oncoprotein. Cell. 2001;104(1):119-30. doi:10.1016/s0092-8674(01)00196-9. PMID: 11163245.
28. Hamdan FF, Srour M, Capo-Chichi JM, Daoud H, Nassif C, Patry L, et al. De novo mutations in moderate or severe intellectual disability. PLoS Genet. 2014;10(10):e1004772. doi:10.1371/journal.pgen.1004772. PMID: 25356899; PMCID: PMC4214635.
29. Deciphering Developmental Disorders Study. Prevalence and architecture of de novo mutations in developmental disorders. Nature. 2017;542(7642):433-8. doi:10.1038/nature21062. PMID: 28135719; PMCID: PMC6016744.
30. Pan X, Liu S, Feng X, Liu L, Zhang X, Qian G, et al. Whole exome sequencing and transcriptome analysis in two unrelated patients with novel SET mutations. J Hum Genet. 2023;68(12):867-74. doi:10.1038/s10038-023-01196-4. PMID: 37737486.
31. Zhang L, Gao J, Liu H, Tian Y, Zhang X, Lei W, et al. Pathogenic variants identified by whole-exome sequencing in 43 patients with epilepsy. Hum Genomics. 2020;14(1):44. doi:10.1186/s40246-020-00294-0. PMID: 33287870; PMCID: PMC7720389.
32. Marinakis NM, Svingou M, Veltra D, Kekou K, Sofocleous C, Tilemis FN, et al. Phenotype-driven variant filtration strategy in exome sequencing toward a high diagnostic yield and identification of 85 novel variants in 400 patients with rare Mendelian disorders. Am J Med Genet A. 2021;185(8):2561-71. doi:10.1002/ajmg.a.62338. PMID: 34008892.
33. Brunet T, Jech R, Brugger M, Kovacs R, Alhaddad B, Leszinski G, et al. De novo variants in neurodevelopmental disorders—experiences from a tertiary care center. Clin Genet. 2021;100(1):14-28. doi:10.1111/cge.13946. PMID: 33619735.
34. Van Vlierberghe P, van Grotel M, Tchinda J, Lee C, Beverloo HB, van der Spek PJ, et al. The recurrent SET-NUP214 fusion as a new HOXA activation mechanism in pediatric T-cell acute lymphoblastic leukemia. Blood. 2008;111(9):4668-80. doi:10.1182/blood-2007-09-111872. PMID: 18299449; PMCID: PMC2343598.
35. Orphanet. Precursor T-cell acute lymphoblastic leukemia (ORPHA:99861). https://www.orpha.net/en/disease/detail/99861
36. Zaghi M, Banfi F, Massimino L, Volpin M, Bellini E, Brusco S, et al. Balanced SET levels favor the correct enhancer repertoire during cell fate acquisition. Nat Commun. 2023;14(1):3212. doi:10.1038/s41467-023-39043-x. PMID: 37270547; PMCID: PMC10239470.
37. National Center for Biotechnology Information. ClinVar: SET gene variants (query SET[gene] AND single_gene[prop], 144 records). https://www.ncbi.nlm.nih.gov/clinvar/?term=SET%5Bgene%5D (accessed 2026-10-28).
38. Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J, et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015;17(5):405-24. doi:10.1038/gim.2015.30. PMID: 25741868; PMCID: PMC4544753.
39. Ehret JK, Engels H, Cremer K, Becker J, Zimmermann JP, Wohlleber E, et al. Microdeletions in 9q33.3-q34.11 in five patients with intellectual disability, microcephaly, and seizures of incomplete penetrance: is STXBP1 not the only causative gene? Mol Cytogenet. 2015;8:72. doi:10.1186/s13039-015-0178-8. PMID: 26421060; PMCID: PMC4587785.
40. Amarillo IE, O'Connor S, Lee CK, Willing M, Wambach JA. De novo 9q gain in an infant with tetralogy of Fallot with absent pulmonary valve: patient report and review of congenital heart disease in 9q duplication syndrome. Am J Med Genet A. 2015;167A(12):2966-74. doi:10.1002/ajmg.a.37296. PMID: 26768185; PMCID: PMC8792801.
41. Bonati MT, Castronovo C, Sironi A, Zimbalatti D, Bestetti I, Crippa M, et al. 9q34.3 microduplications lead to neurodevelopmental disorders through EHMT1 overexpression. Neurogenetics. 2019;20(3):145-54. doi:10.1007/s10048-019-00581-6. PMID: 31209758.
42. Xu X, Zhai Q, Jin H, Yu Y, Han D, Zhang H, et al. SET-NUP214 fusion gene involved early T-cell precursor acute lymphoblastic leukemia in adult with B marker expression. Int J Gen Med. 2021;14:659-64. doi:10.2147/IJGM.S294715. PMID: 33658838; PMCID: PMC7920624.
43. Neul JL, Kaufmann WE, Glaze DG, Christodoulou J, Clarke AJ, Bahi-Buisson N, et al. Rett syndrome: revised diagnostic criteria and nomenclature. Ann Neurol. 2010;68(6):944-50. doi:10.1002/ana.22124. PMID: 21154482; PMCID: PMC3058521.
44. Martin AR, Williams E, Foulger RE, Leigh S, Daugherty LC, Niblock O, et al. PanelApp crowdsources expert knowledge to establish consensus diagnostic gene panels. Nat Genet. 2019;51(11):1560-5. doi:10.1038/s41588-019-0528-2. PMID: 31676867.
45. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. p. 663 (Deletions and duplications—microdeletions and microduplications).
46. Tong Y, Yan S, Shi J, Chen L, Wan C, Xu K. [Genetic analysis of an infant with duplication 9q34 syndrome]. Zhonghua Yi Xue Yi Chuan Xue Za Zhi. 2018;35(3):437-9. doi:10.3760/cma.j.issn.1003-9406.2018.03.030. PMID: 29896750.
47. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. p. 661 (Deletions and duplications—microdeletions and microduplications).
48. GeneGuidance. SPTAN1, STXBP1 and SETBP1 gene records (clinical curation) [database]. Export of 26 September 2026; accessed 28 October 2026.
49. OMIM® Online Mendelian Inheritance in Man. #616078 INTELLECTUAL DEVELOPMENTAL DISORDER, AUTOSOMAL DOMINANT 29; MRD29. Johns Hopkins University; last edited 06/04/2022. https://omim.org/entry/616078. Accessed 28 October 2026.
50. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Schinzel-Giedion syndrome. p. 306.
51. OMIM® Online Mendelian Inheritance in Man. #269150 SCHINZEL-GIEDION MIDFACE RETRACTION SYNDROME. Johns Hopkins University; last edited 21/04/2026. https://omim.org/entry/269150. Accessed 28 October 2026.
52. OMIM® Online Mendelian Inheritance in Man. #180849 RUBINSTEIN-TAYBI SYNDROME 1; RSTS1. Johns Hopkins University; last edited 13/08/2025. https://omim.org/entry/180849. Accessed 28 October 2026.
53. OMIM® Online Mendelian Inheritance in Man. #613684 RUBINSTEIN-TAYBI SYNDROME 2; RSTS2. Johns Hopkins University; last edited 21/07/2025. https://omim.org/entry/613684. Accessed 28 October 2026.
54. OMIM® Online Mendelian Inheritance in Man. #618332 MENKE-HENNEKAM SYNDROME 1; MKHK1. Johns Hopkins University; last edited 01/07/2019. https://omim.org/entry/618332. Accessed 28 October 2026.
55. OMIM® Online Mendelian Inheritance in Man. #618333 MENKE-HENNEKAM SYNDROME 2; MKHK2. Johns Hopkins University; last edited 15/03/2019. https://omim.org/entry/618333. Accessed 28 October 2026.
56. OMIM® Online Mendelian Inheritance in Man. #122470 CORNELIA DE LANGE SYNDROME 1; CDLS1. Johns Hopkins University; last edited 24/04/2026. https://omim.org/entry/122470. Accessed 28 October 2026.
57. European Medicines Agency. Slenyto (melatonin) prolonged-release tablets: EPAR — product information (summary of product characteristics), sections 4.1 and 4.2. EMEA/H/C/004425; first authorisation 20 Sep 2018; extension to neurogenetic disorders (II/0025) 26 Aug 2024; ADHD indication (II/0028) 17 Mar 2025; product information updated 10 Jun 2026 [accessed 7 Nov 2026]. Available from: https://www.ema.europa.eu/en/medicines/human/EPAR/slenyto
58. Janssen Pharmaceuticals. RISPERDAL (risperidone) tablets, orally disintegrating tablets, oral solution. Prescribing information. DailyMed setid 7e117c7e-02fc-4343-92a1-230061dfc5e0 (published 26/10/2026). https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=7e117c7e-02fc-4343-92a1-230061dfc5e0
59. European Medicines Agency. Risperdal — Article 30 referral: Annex I–IV (harmonised summary of product characteristics, sections 4.1 and 4.2); CHMP opinion of 24 Jul 2008 and European Commission decision of 7 Oct 2008. Available from: https://www.ema.europa.eu/en/medicines/human/referrals/risperdal
60. Sandoz/Novartis. RITALIN (methylphenidate hydrochloride) tablets. Prescribing information (NDA010187). DailyMed setid d6fb2750-cdab-4749-ba0d-7534840a5892 (21/02/2025). https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=d6fb2750-cdab-4749-ba0d-7534840a5892
61. UCB, Inc. KEPPRA (levetiracetam) tablets, oral solution; KEPPRA injection. Prescribing information (NDA021035, NDA021505). DailyMed setid 3ca9df05-a506-4ec8-a4fe-320f1219ab21 (version of 28/01/2026). https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=3ca9df05-a506-4ec8-a4fe-320f1219ab21 (accessed 2026-10-28).