Genetic testing and analysis of 1024 children with global developmental delay or intellectual disability: a single-center cohort study Authors: Bing Wang, Xinna Ji, Fan Wu, Mengyao Shen, Ping Zheng, Shuo Feng, Huanhuan Wu, Shupin Li, Aijie Liu, Lina Xie, Xue Zhang, Qian Chen Published: 2026 ABSTRACT AbstractGlobal developmental delay (GDD) and intellectual disability (ID) are common neurodevelopmental disorders with a strong genetic basis. However, comprehensive large-cohort analyses integrating genotype–phenotype correlations and functional mechanisms remain limited. This study aimed to systematically characterize the clinical and genetic spectrum of GDD/ID in a large single-center cohort and to explore the functional attributes of disease-causing genes. We retrospectively analyzed 1024 children diagnosed with GDD or ID who underwent genetic testing, including trio whole-exome sequencing (trio-WES), proband-only WES, and clinical exome sequencing. Clinical phenotypes were categorized, and functional enrichment analyses were conducted for genes associated with diagnostic and probable diagnostic results. A genetic diagnosis was achieved in 48.1% of patients, with trio-WES demonstrating a significantly higher diagnostic yield than proband-only approaches. Pathogenic variants mainly comprised single-nucleotide variants and copy number variants. Identified genes were predominantly involved in protein homeostasis, synaptic and ion channel function, epigenetic regulation, and key developmental signaling pathways. Distinct genotype–phenotype associations were observed among clinical subgroups, including enrichment of synaptic-related genes in epilepsy-associated GDD/ID and epigenetic regulatory genes in patients with facial dysmorphism. Conclusion: This study provides a comprehensive characterization of the genetic landscape of GDD/ID in a large single-center cohort and identifies distinct genotype–phenotype correlations and convergent molecular pathways underlying these disorders. What is Known:• Global developmental delay (GDD) and intellectual disability (ID) are highly heterogeneous neurodevelopmental disorders with a strong genetic basis.What is New:• We report a large single-center cohort of 1024 children with GDD/ID, providing a comprehensive overview of the genetic landscape and diagnostic yield of different sequencing strategies.• Our study systematically integrates genotype–phenotype correlations with functional pathway analyses, highlighting key molecular mechanisms underlying GDD/ID and supporting refined molecular stratification.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00431-026-07111-1.Keywords: Neurodevelopmental disorders, Global developmental delay, Intellectual disability, Whole-exome sequencing, Genetic diagnosis, Genotype–phenotype correlation, Functional pathway analysis Introduction Global Developmental Delay (GDD) and Intellectual Disability (ID) are among the most common neurodevelopmental disorders. GDD is characterized by pathological delays in at least two developmental domains, including gross/fine motor skills, language, cognition, social interaction, and adaptive functions, in children under 5 years of age. ID refers to significant impairments in intellectual and social adaptive functioning in children over 5 years of age[1]. The global prevalence of GDD/ID is estimated to be between 1.3% and 1.4%. These conditions not only pose a serious threat to children's health but also bring substantial medical, psychological, economic, and social burdens[2]. Early diagnosis, effective treatment, and prognosis evaluation are crucial for alleviating the national public health burden. Genetic factors are an important cause of GDD/ID, accounting for 40–60% of all cases in various research cohorts[3]. To date, 1882 genes have been definitely linked to GDD/ID in the OMIM database. However, precise intervention strategies for GDD/ID are still lacking. Therefore, genetic testing to better understand the etiology of GDD/ID and identify innovative intervention targets is essential for improving management and prognosis. The genetic variations associated with GDD/ID are diverse, including chromosomal abnormalities, copy number variations (CNVs), single nucleotide variations (SNVs), small insertions/deletions (indels), as well as mitochondrial variants, mosaicism, and epigenetic abnormalitie[4]. Clinical exome testing, also known as gene panel testing, is an efficient and cost-effective choice for patients with classic clinical presentations highly suspicious of specific monogenic disorders, while Whole Exome Sequencing (WES) can provide a more comprehensive search for genetic causes in complex or atypical cases. In 2019, the American College of Medical Genetics and Genomics (ACMG) recommended WES as a first-line diagnostic tool for children with GDD/ID[5]. However, limitations persist: there is a lack of large-scale, systematic genetic profiles for GDD/ID patients, and the genotype–phenotype associations for these conditions remain poorly defined. Furthermore, there is a shortage of integrative functional analyses at the molecular pathway level to better understand the pathogenic mechanisms. In light of these gaps, we established a cohort of 1024 GDD/ID patients from a single center. We hypothesize that integrated genomic and phenotypic analyses will define the genetic architecture of GDD/ID, establish genotype–phenotype correlations, and identify convergence of pathogenic variants on core neurodevelopmental pathways. Accordingly, this study systematically integrates detailed clinical phenotyping with genomic data (from WES/CES) to characterize the genetic landscape and uncover shared pathogenic pathways. Materials and methods Study population This retrospective cohort study included patients diagnosed with Global Developmental Delay (GDD) or Intellectual Disability (ID) at Capital Center for Children's Health, Capital Medical University in China, from January 2015 to December 2024, who underwent genetic testing. The inclusion criteria were: (1) Diagnosis of GDD/ID based on the DSM-5 criteria for GDD/ID [1]. GDD refers to children under 5 years of age with pathological delays in at least two of the five developmental domains—gross/fine motor skills, language, cognition, social interaction, and adaptive functions, or failure to meet developmental milestones for their age. ID refers to children over 5 years with significant deficits in intellectual and adaptive functioning during the developmental period. (2) The Gesell Developmental Scale score lower than 75 in at least two domains for children under 5 years, or the IQ score under 70 on the Wechsler Intelligence Scale for Children for children aged 5 years or older. (3) Written informed consent for genetic testing from the legal guardians. (4) Exclusion of cases where GDD/ID was caused by clear etiologies such as head trauma, brain tumors, CNS infections, cerebrovascular diseases, other acquired brain injuries, or numerical chromosomal abnormalities. All patients were consecutively enrolled in this study, including those who had undergone prior genetic testing at other institutions. All enrolled children were evaluated by a multidisciplinary team comprising two pediatric neurologists and a genetic analyst, and genetic testing was performed in all individuals who met the predefined inclusion and exclusion criteria. The study was approved by the Institutional Review Board of the Capital Institute of Pediatrics (IRB:SHERLL2022064) and was conducted in accordance with the Declaration of Helsinki. Written consent was obtained from all parents or legal guardians. The genetic testing results were provided to patients, their families, and treating physicians as part of the diagnostic, therapeutic, and care reference. Genetic Testing Analysis A total of 1024 patients underwent genetic testing, including 715 who received trio whole-exome sequencing (trio-WES), 90 who underwent proband-only WES, and 219 who underwent clinical exome sequencing (CES); among these, 200 patients additionally underwent mitochondrial gene sequencing. Genetic testing strategies evolved over time with advances in sequencing technologies: targeted clinical exome sequencing (CES) was predominantly used between 2015 and 2016, whereas whole-exome sequencing (WES) became more widely adopted after 2017. The choice of testing approach, including trio-WES, was determined by parental sample availability and technical considerations, rather than predefined clinical severity or syndromic features. Trio-WES involves analyzing the proband and their biological parents as a single unit, while proband-only WES refers to sequencing only the proband. CES is based on clinical phenotypes (GDD/ID, epilepsy, ASD, etc.) and focuses on high-throughput sequencing of clinically relevant exonic regions. CNV detection methods included NGS-based CNV calling and single-nucleotide polymorphism (SNP) microarray analysis. Sample processing, DNA sequencing, and variant detection methods are detailed in the supplementary materials. The pathogenicity of variants was classified according to the American College of Medical Genetics (ACMG) guidelines, with categories including pathogenic (P), likely pathogenic (LP), variants of uncertain significance (VUS), likely benign (LB), and benign (B)[6]. The pathogenicity classification of CNV was performed according to the joint consensus standards published by the ACMG and the Clinical Genome Resource (ClinGen) in 2020[7]. Given that pathogenicity classification may change as evidence accumulates, we reassessed the pathogenicity of all variants in June 2025 to ensure classification consistency. Diagnostic results were defined as: (1) pathogenic or likely pathogenic variants in autosomal dominant genes; (2) pathogenic or likely pathogenic variants in X-linked genes; (3) two pathogenic or likely pathogenic variants in both alleles of autosomal recessive genes; (4) pathogenic or likely pathogenic CNVs. Possibly diagnostic results were defined as: (1) VUS variants matching autosomal dominant or X-linked inheritance; (2) two VUS variants in autosomal recessive genes; (3) one pathogenic/likely pathogenic and one VUS variant in autosomal recessive genes; (4) CNVs classified as VUS. Negative results were defined as the absence of pathogenic/likely pathogenic or VUS variants, or a single VUS variant in both alleles of autosomal recessive genes. For female carriers of X-linked recessive variants, the diagnostic classification was defined as follows: carriers with pathogenic or likely pathogenic (P/LP) variants who also presented with corresponding phenotypes were considered positively diagnosed. Asymptomatic carriers of P/LP variants, as well as all carriers of variants of uncertain significance (VUS), were considered negative. Functional clustering of genes We performed functional enrichment analysis using the Metascape online platform (https://metascape.org/). Genes included in the analysis were those associated with diagnostic results and possibly diagnostic results. Enrichment analyses were conducted for Gene Ontology (GO) terms (biological process, molecular function, and cellular component) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Homo sapiens genome-wide genes were used as the background gene set. Statistical significance was defined as a p-value < 0.01, and multiple testing correction was performed using the Benjamini–Hochberg method, with a false discovery rate (FDR) < 0.05 considered significant. Data integrity The inclusion criteria for this study required that patients have complete clinical assessment records and genetic testing results. Therefore, the final dataset used for analysis (N = 1,024) has no missing values for the predefined clinical phenotype and genetic results, and all patients were included in the full primary data analysis. Clinical data were retrospectively extracted from the electronic medical record system using a pre-designed standardized data collection form. To ensure data quality, all extracted variables were cross-verified by two investigators, and a random 10% sample was manually checked against the original records. Statistical analysis Categorical variables were compared using the Chi-square test or Fisher’s exact test, as appropriate. Continuous variables were compared using the Student’s t-test or the Kruskal–Wallis test, depending on their distribution. To evaluate factors associated with diagnostic yield, univariate logistic regression analyses were first performed. Variables with a P value < 0.05 in univariate analysis were included in the multivariate logistic regression model. Multicollinearity among independent variables included in the multivariate logistic regression model was assessed using variance inflation factors (VIF). All included variables showed VIF values close to 1, indicating no significant multicollinearity. Given that some patients could present with multiple phenotype subgroups, each clinical feature was treated as an independent binary variable (present/absent) in the regression model. The unit of analysis was the individual patient; therefore, the assumption of independence of observations was preserved. Model assumptions for logistic regression, including linearity in the logit for continuous variables and absence of influential outliers, were evaluated and considered to be satisfied. All statistical analyses were performed using SPSS software (version 31.0). For multiple comparisons, P values were adjusted using the Bonferroni correction, with a corrected P < 0.05 considered statistically significant. Data visualization was performed using GraphPad Prism, and additional analyses were conducted using R (version 4.5.1). Artificial intelligence statement No artificial intelligence (AI)-assisted tools were used for data analysis or interpretation. Results Demographic and clinical data of the study population This study enrolled a total of 1024 patients diagnosed with Global Developmental Delay (GDD) or Intellectual Disability (ID), including 632 males and 392 females, with a male-to-female ratio of 1.6:1. The median age at the time of genetic testing was 2.7 years, with an interquartile range of 1.2–5.1 years. Among the 1,024 patients, 493 tested positive for genetic findings, resulting in a diagnostic rate of 48.1% (95% CI: 45.0%−51.2%). The diagnostic rate for females was significantly higher than that for males (P = 0.002). A relatively large proportion of the tested population consisted of infants and toddlers (54.2%), with the diagnostic rate independent of age. Among the 1024 patients, 715 (69.8%) underwent trio-WES, 90 (8.8%) underwent proband-only WES, and 219 (21.4%) had clinical exome sequencing (CES). Controlling for baseline phenotypic characteristics (Supplementary Table 2), the diagnostic rate for trio-WES (51.3%) was significantly higher than that for proband-only WES and CES (40.8%) (P = 0.002). Based on clinical phenotypes, the patients were divided into two main groups: the non-syndromic GDD/ID group (n = 306, 29.9%) and the syndromic GDD/ID group (n = 718, 70.1%). To further explore the genotype–phenotype relationship, the syndromic GDD/ID group was divided into four subgroups: GDD/ID + epilepsy (n = 322, 31.4%), GDD/ID + ASD (n = 77, 7.5%), GDD/ID + facial dysmorphism (n = 163, 15.9%), and GDD/ID + muscle tone abnormalities (n = 173, 16.9%). Based on clinical features, the same patient may belong to multiple subgroups. Among all clinical phenotypes, the GDD/ID + epilepsy group had the highest proportion (31.4%), with focal seizures in 137 cases and generalized seizures in 96 cases. The diagnostic rate was significantly higher in the GDD/ID + facial dysmorphism group (57%, P = 0.013) (Table 1). Table 1Demographic characteristics of patientsVariablesNo. of patientsNo. of patients with diagnostic resultsUnivariate analysismultivariate analysisOR(95%CI)P valueOR(95%CI)P valueTotal1024493(48.1%)95% CI:45.1%−51.2%SexFemale392213(54.3%)0.668(0.519–0.861)0.0020.698(0.540–0.903)0.006Male632280(44.3%)AgeInfancy (0–1 year)235117(49.8%)0.98(0.944–1.017)0.292Toddler (1–3 years)320163(50.9%)Early childhood (3–6 years)273120(44.0%)Middle childhood (6–12 years)16883(49.4%)Adolescent (12–18 years)2810(35.7%)StrategiesTrio-WES715367(51.3%)0.8(0.688–0.931)0.0020.607(0.387–0.953)0.030Proband-only-WES9036(40%)CES21990(41.1%)Non-syndromic/Syndromic*Syndromic GDD/ID718343(47.8%)0.951(0.728–1.243)0.714Non-syndromic GDD/ID306150(49%)EpilepsyYes322159(49.4%)1.075(0.825–1.399)0.592No702334(47.6%)Facial DysmorphismYes16393(57.1%)1.511(1.077–2.119)0.0171.475(1.046–2.080)0.027No861400(46.5%)Muscle tone abnormalitiesYes17386(49.7%)1.078(0.778–1.495)0.651No851407(47.8%)ASDYes7721(27.3%)0.377(0.225–0.633) < 0.0010.389(0.231–0.656) < 0.001No947472(49.8%)Family history**Yes5528(50.9%)1.124(0.653–1.936)0.673No969465(48.0%)*Non-syndromic GDD/ID: Intellectual disability is the sole clinical manifestation. Syndromic GDD/ID: Intellectual disability is accompanied by one or more clinical phenotypes, such as epilepsy, macrocephaly, microcephaly, hypotonia, dysmorphic features, cardiac malformations, etc.**Relatives with NDD or epilepsy or FC or neurofibroma within three generations Demographic characteristics of patients *Non-syndromic GDD/ID: Intellectual disability is the sole clinical manifestation. Syndromic GDD/ID: Intellectual disability is accompanied by one or more clinical phenotypes, such as epilepsy, macrocephaly, microcephaly, hypotonia, dysmorphic features, cardiac malformations, etc. **Relatives with NDD or epilepsy or FC or neurofibroma within three generations Analysis of diagnosis results Overall, 493 patients received a diagnostic result (Fig. 1), resulting in a diagnostic yield of 48.1% (95% CI: 45.0%−51.2%) (Supplementary Table 1). Among the 1024 enrolled patients, 60 had previously undergone genetic testing at other institutions before referral to our center. Within this subgroup, 29 patients had diagnostic findings, 6 had possibly diagnostic findings, and 25 had negative results. After excluding these previously tested patients with known molecular results, the diagnostic yield in the remaining cohort was 48.1%, which was essentially unchanged from that observed in the overall cohort. Flowchart of Study Design (This flowchart includes only patients who underwent genetic testing; the total number of initially assessed or excluded cases was not available.) Specifically, among the 1024 enrolled patients, 365 cases (74.0%) carried only SNV/indel variants, 113 cases (22.9%) carried only CNV variants, and 15 cases (3.0%) harbored both SNV/indel and CNV variants(Fig. 2A). Notably, among these 15 patients, 7 cases exhibited SNV/indel and CNV variants forming compound heterozygous configurations in autosomal recessive disorders. A total of 15 patients had dual or triple genetic diagnoses, including six patients with dual diagnoses involving SNV/indel and CNV, two patients with triple diagnoses involving SNV/indel and CNV, four patients with dual CNV diagnoses, and three patients with dual SNV/indel diagnoses. Re-analysis of variant classification in June 2025 identified 12 additional patients with pathogenic or likely pathogenic variants, resulting in an increase in the overall diagnostic yield from 47.0% to 48.1%. SNV/indel variations involved 440 variants in 223 genes. 136 CNVs were detected in this cohort. Some CNVs were recurrent while others were unique. In total, 98 different types of CNVs were identified. Spectrum and characteristics of diagnostic variants in the GDD/ID cohort, including variant category (A), modes of inheritance (B), parental origin (C), and molecular variant types (D) Among the 380 patients with SNV/indel variations, 66.1% (251/380) exhibited autosomal dominant disorders, 16.3% (62/380) exhibited autosomal recessive disorders, 15.8% (60/380) exhibited X-linked disorders, and 1.8% (7/380) had mitochondrial variants (Fig. 2B). Of the 440 SNV/indel variations, 426 had parental verification data, with 60.3% (257/426) being de novo variants and 39.7% (169/426) inherited from parents (Fig. 2C). Of these, 210 were missense variants, 88 were nonsense variants, 89 were frameshift variants, 34 were splice variants, 10 were deletions, and 9 were other types of variants, including 3 repeat expansions, 2 insertions, 2 mitochondrial tRNA gene variants, 1 synonymous variant, and 1 variant in the start codon (Fig. 2D). The 12 genes contributing most to the diagnostic results (≥ 5 cases) are listed in Supplementary Table 3, and the genotype–phenotype relationships are shown in Fig. 3A. The most frequently affected genes were MECP2 (13/1024, 1.2%) and SYNGAP1 (10/1024, 1%). In non-syndromic GDD/ID, MECP2 (6/306, 2.0%) and MMACHC (6/306, 2.0%) variants were most prevalent, while SYNGAP1, SCN2A, STXBP1, and SLC2A1 were more common in the GDD/ID + epilepsy group. Heatmap of identified recurrent genes with diagnostic SNV/indel(A) and CNVs(B) among different phenotype subgroups Among the 136 CNV loci, the CNV size ranged from 53 to 140,643,621 bp, with an average size of 7.45 Mb. Among them, 97 cases (74.0%) had deletions, and 39 cases (29.8%) had duplications. 3 cases had uniparental disomy (UPD), including 1 case of maternal 15q11.2q26.3 UPD with a phenotype of Prader-Willi syndrome and 2 cases of paternal 15q11.2q26.3 UPD with a phenotype of Angelman syndrome. The top three most frequent CNV loci were 15q11.2q13.1 del (13/136), 7q11.23 del (6/136), and 16p11.2 del (4/136). The genotype–phenotype relationships for common CNV variants (≥ 3 cases) are shown in Fig. 3B, and the phenotypes associated with 15q11.2q13.1 del include a wide range of features, such as GDD/ID, epilepsy, facial dysmorphism, and abnormal muscle tone. Notably, non-syndromic GDD/ID was observed in carriers of all the common CNVs listed in Fig. 3B. Among the 493 cases with positive diagnoses, the most frequent chromosomal locations of the variants were X chromosome (72/576, 12.5%), chromosome 1 (55/576, 9.5%), chromosome 2 (41/576, 7.1%), and chromosome 15 (41/576, 7.1%) (Supplementary Fig. 1). Genes with hotspot variants (≥ 3 cases) on each chromosome are listed in Supplementary Table 4. In this cohort, mitochondrial DNA sequencing was performed in 200 probands, among whom pathogenic mitochondrial variants were identified in 7 individuals. Two patients harbored variants in MT-ATP6: one maternally inherited variant m.9185 T > C(heteroplasmy 100%, maternal heteroplasmy 34.61%), and one de novo variant m.8993 T > C (heteroplasmy 100%, not detected in the mother). Two patients carried MT-ND3 variants m.10197G > A (heteroplasmy 100%, maternal heteroplasmy 32.5%) and m.10158 T > C (heteroplasmy 67.5%, maternal heteroplasmy 5.13%). One patient carried an MT-TI variant m.4296G > A (heteroplasmy 99.24%, maternal heteroplasmy 90.28%). These three variants were maternally inherited. In addition, one patient carried a de novo MT-ND5 variant m.13513G > A (heteroplasmy 60.79%, not detected in the mother), and another patient carried a de novo MT-TL1 variant m.3243A > G (heteroplasmy 88.0%, not detected in the mother). None of the mothers of these 7 patients exhibited neurodevelopmental disorder–related phenotypes and additional phenotypes commonly associated with high levels of mitochondrial heteroplasmy. Among the 493 patients with positive genetic diagnoses, mosaic variants were identified in 5 probands, with variant allele fractions ranging from 25 to 53%. All mosaic variants were classified as de novo, although the possibility of low-level parental germline mosaicism could not be completely excluded. Three of these mosaic variants were located on the X chromosome (3/5, 60%). In addition, mosaic pathogenic variants were detected in the mothers of 6 probands, with variant allele fractions ranging from 6 to 50%; half of these variants were located on the X chromosome (3/6, 50%). All carrier mothers were clinically unaffected. Notably, in one patient with GDD/ID accompanied by epilepsy carrying a KRAS variant (NM_004985.3:c.35G > A, p.G12D), the variant was not detected in peripheral blood samples from either the proband or the parents. However, somatic mosaicism of this variant (33%) was identified in resected tissue from a facial sebaceous nevus. This variant is associated with Schimmelpenning syndrome, a neurocutaneous disorder characterized by sebaceous nevi in combination with brain, ocular, or skeletal abnormalities, with neurological involvement typically manifesting as epilepsy and developmental delay. The molecular findings were highly consistent with the patient’s clinical phenotype. Analysis of possibly diagnosis results Among the 113 probands with possibly diagnostic results (11.1%), a total of 148 variants of uncertain significance (VUS) were identified, including 123 SNV/indel variants and 24 CNV events. Among patients harboring SNV/indel variants, autosomal recessive disorders were the most common inheritance pattern (54.8%, 51/93), and missense variants accounted for the majority of all VUS (114/148, 77.0%). Of the 24 CNVs classified as VUS, 8 (27.3%) were duplications, 11 (45.8%) were deletions, and 7 (29.2%) represented uniparental disomy (UPD). Notably, the following genes were each identified in two probands within the potentially diagnostic category: ARHGEF9, BCORL1, COQ4, GABRA1, NARS2, NPC1, PLA2G6, TH, KDM5C, and ZNF292. These findings highlight a subset of recurrent candidate genes that may be associated with GDD/ID. However, further functional studies and the accumulation of additional well-characterized cases are required to provide sufficient evidence for reclassification of these VUS and to refine their pathogenicity assessments. Functional and pathway analysis of GDD/ID-associated genes A total of 335 genes associated with diagnosis results and possibly diagnosis results were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation and enrichment analyses (Supplementary Tables 5–6). Based on their primary biological functions, those genes were broadly classified into four major functional categories: protein homeostasis (161 genes), synapses and ion channels (101 genes), epigenetic regulation (77 genes), and signaling pathways (53 genes) (Fig. 4; Supplementary Tables 7–11). A subset of genes was shared across two or more functional categories, reflecting functional overlap and pleiotropy (Supplementary Table 12). Venn diagram of functional classification of diagnosed positive and possibly positive genes of GDD/ID Comparative analyses of functional clustering and clinical phenotypes revealed distinct genotype–phenotype associations (Fig. 5). Genes involved in protein homeostasis were most frequently observed in patients with nonsyndromic GDD/ID. In contrast, genes related to synaptic function and ion channels were predominantly enriched in patients with GDD/ID accompanied by epilepsy, whereas genes associated with epigenetic regulation were relatively overrepresented in patients with facial dysmorphism. The numerical distribution of clinical phenotype subgroups among GDD/ID genes of different functional classification Genes classified within the signaling pathway category were mapped to 23 distinct pathways, which are involved in key biological processes including cell proliferation, differentiation, metabolism, stress responses, and immune and inflammatory regulation. Enrichment analysis demonstrated that the Ras–MAPK and PI3K/AKT–mTOR pathways were the two most significantly enriched signaling pathways associated with GDD/ID–related genes. The Hippo and Wnt signaling pathways were also prominently represented. Across all signaling pathways, the most frequently implicated genes were KRAS, MAP2K1, and MTOR (each identified in ≥ 10 cases) (Supplementary Fig. 2), underscoring their central roles as hub genes within neurodevelopmental signaling networks. Discussion In this study, a total of 1024 children with GDD/ID underwent genetic analysis, yielding an overall diagnostic rate of 48.1%. A meta-analysis by Karen Y et al., encompassing 37 studies worldwide, reported an overall diagnostic yield of 42% (95% CI, 35%–50%) for whole-exome sequencing (WES) in patients with ID/GDD, with a significantly higher diagnostic yield observed in Chinese cohorts compared with Western populations (55% vs 35%)[8]. Recent single-center and multicenter studies from China have reported diagnostic yields ranging from 46.1% to 61%, including rates of 46.1%, 61%, 42.6%, and 49.7%[9]. The diagnostic yield observed in the present study appears to be broadly consistent with previously published data.The increase in diagnostic yield after variant reclassification (47.0% to 48.1%) was driven by updated evidence supporting the pathogenicity of previously uncertain variants, highlighting the importance of periodic re-evaluation in genetic testing. The diagnostic yield of trio-WES was significantly higher than that of proband-only WES or targeted gene panels, a finding that has been consistently reported in prior studies[9, 12]. This difference may be partly attributable to the prominent contribution of de novo variants to the genetic architecture of ID, which accounted for 60.3% of pathogenic variants in our cohort. Trio-WES improves the detection sensitivity for de novo variants and enables more effective exclusion of benign variants inherited from unaffected parents. In this study, the proportion of male patients was higher than that of females. This difference can be partly attributed to X-linked genetic defects, as X-linked recessive inheritance increases the disease risk in males[11]. It may also be explained by the “female protective effect” proposed by Jacquemont et al., which suggests that males are less tolerant to variants and are more likely to manifest phenotypes at the same genetic burden, whereas females may require a higher mutational load to develop symptoms. This may explain why the diagnostic yield in females was significantly higher than in males in our cohort[13]. Notably, patients with GDD/ID accompanied by facial dysmorphism exhibited a significantly higher diagnostic yield, consistent with previous reports [10, 11, 14]. These findings suggest that careful clinical evaluation to identify facial dysmorphism may be an important step, as patients with such features may benefit from genetic testing to facilitate etiologic diagnosis. Analysis of the pathogenic variant spectrum revealed that the five most frequently affected genes in this cohort were MECP2 (12 cases), SYNGAP1 (10 cases), SCN2A (9 cases), MMACHC (9 cases), and STXBP1 (7 cases). Among these, MECP2, SCN2A, and STXBP1 have been consistently identified as high-frequency pathogenic genes across multiple large-scale cohorts[15, 16]. In contrast, SYNGAP1 and MMACHC were more prominently represented in the present cohort, suggesting that genetic spectra may vary across different research populations. Previous studies have also identified SCN1A and KCNQ2 as recurrently mutated genes in ID. Furthermore, a large Chinese CMA study involving 10 026 children with ID reported the eight most common pathogenic genomic regions to be 15q11–q13, 7q11.23, 22q11.2, 1p36, 22q13.3, 16p11.2, and 4p16.3[15]. The top three CNV variants in our study are included in these regions, supporting the consistency of our findings with previous reports. In this cohort, mosaic variants were identified in 5 probands and maternal mosaicism was detected in 6 mothers. Notably, 55% of mosaic variants were located on the X chromosome, suggesting a potential association with the mechanism of X-chromosome inactivation. X-chromosome inactivation involves large-scale epigenetic remodeling, which may increase genomic instability and thereby predispose the X chromosome to mosaic variation[17].Although the overall proportion of maternal mosaicism was relatively low in this cohort—approximately two-thirds of maternal mosaic variants exhibited variant allele fractions below 20%—the mothers were clinically unaffected, whereas their offspring manifested disease-related phenotypes. These observations indicate that even low-level maternal mosaicism in somatic tissues may coexist with a higher mutational burden in germ cells, which may increase the genetic risk and disease susceptibility in offspring.In 2012, Jeong Ho Lee and Annapurna Poduri first reported brain-specific somatic mosaic variants in PIK3CA, AKT3, and MTOR in patients with hemimegalencephaly (HMG) [18, 19]. In the present cohort, one patient with Schimmelpenning syndrome carrying a KRAS variant did not undergo direct molecular analysis of brain tissue; however, the patient’s clinical features were highly consistent with this diagnosis. We therefore hypothesize the presence of a pathogenic somatic mosaic variant in neural tissue in this case.These findings suggest that a negative result from peripheral blood–based genetic testing does not fully exclude the presence of tissue-specific somatic mosaicism. When clinical features are highly suggestive of a defined genetic syndrome, targeted genetic testing of affected tissues may be essential for establishing an accurate molecular diagnosis. In this study, 48.1% of patients remained undiagnosed despite comprehensive genetic testing. Several factors may account for this. First, current exome-based approaches have limited sensitivity for detecting deep intronic variants, regulatory non-coding variants, and low-level mosaicism. Second, structural variants and complex genomic rearrangements may not be reliably detected by WES or CES due to inherent technical limitations. Third, some patients may harbor pathogenic variants in genes not yet associated with disease or in regions with incomplete annotation. Finally, a subset of cases may be attributable to non-genetic or multifactorial etiologies. In this context, genome sequencing may provide additional diagnostic value, particularly for variants not detectable by exome-based approaches. In addition, further functional characterization of NDD-associated genes may improve our understanding of gene–disease relationships and facilitate the identification of additional pathogenic variants. Human brain development is achieved through precise spatiotemporal genetic regulation[20]. In neurodevelopmental disorders, multiple molecular mechanisms often contribute to disease onset. Functional enrichment analysis in this study shows that pathogenic genes predominantly cluster into three functional categories: synapses and ion channels, protein homeostasis, and epigenetic regulation. These pathways interact with one another to influence the formation and maturation of neural networks [21, 22]. It is noteworthy that genes associated with epigenetic regulation are more frequently observed in patients with GDD/ID and facial dysmorphism. This suggests that epigenetic mechanisms play a crucial role in early embryonic morphogenesis. The migration, proliferation, and differentiation of neural crest cells are tightly regulated by epigenetic signals, such as DNA methylation and histone modifications, in a spatiotemporal manner. As a result, variants in these genes may affect both neural and facial development, contributing to a syndromic phenotype. The Ras-MAPK and PI3K-AKT-mTOR signaling cascades are critical pathways involved in the regulation of cell growth, proliferation, and differentiation[23]. These pathways influence brain development and synaptic plasticity by modulating gene transcription and protein synthesis at downstream levels. Notably, a significant number of pathogenic genes associated with neurodevelopmental disorders are also key drivers of cancer, both sharing the Ras-MAPK and PI3K-AKT-mTOR signaling pathways but resulting in markedly different clinical phenotypes[24]. For example, the KRAS gene, a well-known oncogene closely linked to colorectal cancer, lung cancer, and pancreatic cancer[25, 26], presents variants (KRAS c.65A > G and c.65A > G) in this cohort that were only associated with intellectual disability (ID) and epilepsy. This discrepancy is primarily attributed to differences in the timing and intensity of the variant, as well as dynamic regulation by chromatin remodeling complexes and the immune microenvironment. In general, variants associated with neurodevelopmental disorders typically arise in the germline or during embryonic development, often with relatively weaker intensity, affecting neuronal differentiation and migration during early brain development. In contrast, cancer-related variants are often somatic, strongly potent variants that drive uncontrolled proliferation later in life[24, 27]. Although this is a single-center study, several limitations should be acknowledged. First, the total number of patients initially assessed for GDD/ID and the number of individuals excluded according to predefined criteria were not systematically recorded and are therefore unavailable, which may limit the transparency of cohort selection. Second, although a small proportion of patients had received prior genetic testing at other institutions, exclusion of these cases did not materially alter the overall diagnostic yield, supporting the robustness of our findings. In addition, we acknowledge that clinical features suggestive of somatic mosaicism (such as segmental pigmentation abnormalities or body asymmetry) were not systematically assessed in all patients, and mosaicism detection primarily relied on peripheral blood–based testing, which may have led to under-recognition of tissue-restricted or subtle mosaic cases. Finally, while the diagnostic rate and primary pathogenic gene spectrum observed in this cohort are broadly consistent with those reported in previous international studies, differences in genetic background, referral patterns, and testing strategies across populations may still influence the observed spectrum. Therefore, further multicenter studies involving diverse populations are needed to validate the generalizability of these findings. This study supports the potential value of integrating genetic testing into the clinical evaluation of GDD/ID and suggests that incorporating WES into patient care may provide useful information for diagnosis and clinical management. Conclusion This study provides a comprehensive characterization of the genetic landscape of GDD/ID in a large single-center cohort. Distinct genotype–phenotype relationships were observed, and pathogenic variants converged on key biological categories, including synaptic function and ion channels, protein homeostasis, and epigenetic regulation. These findings provide a systematic overview of the genetic landscape underlying GDD/ID and may facilitate further understanding of disease mechanisms. Supplementary Information Below is the link to the electronic supplementary material. (DOCX 636 KB) (DOCX 605 KB) Acknowledgements This work was supported by the National Key Research and Development Program of China(No. 2022YFC2703903. Authors’ contributions Conceptualization: Bing Wang, Xinna Ji, and Qian Chen; Methodology: Bing Wang, Xinna Ji, and Fan Wu; Formal Analysis: Bing Wang and Fan Wu; Investigation: Xinna Ji, Fan Wu, Mengyao Shen, Ping Zheng, Shuo Feng, Huanhuan Wu, Shupin Li, Aijie Liu, and Lina Xie; Writing – Original Draft Preparation: Bing Wang, Xinna Ji, and Fan Wu; Writing – Review & Editing: Xinna Ji and Qian Chen; Supervision: Xinna Ji, Qian Chen, and Xue Zhang; Funding Acquisition: Xue Zhang and Qian Chen. All authors have read and approved the final manuscript. Funding This work was funded by National Key Research and Development Program of China (No. 2022YFC2703903). Data Availability Data supporting the findings of this study are provided in the article and supplementary materials. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials (DOCX 636 KB) (DOCX 605 KB) Data Availability Statement Data supporting the findings of this study are provided in the article and supplementary materials.