Clinical utility and genetic landscape of exome sequencing in a large pediatric epilepsy cohort: Insights from a Turkish tertiary care center

Abstract

Objective

To evaluate the diagnostic utility and genetic spectrum of next‐generation sequencing (NGS) in a large, well‐phenotyped cohort of Turkish pediatric patients with epilepsy of unknown etiology.

Methods

Between January 2021 and December 2024, 250 children (115 female, 135 male) with unexplained epilepsy underwent either whole‐exome sequencing (WES; n = 104) or clinical exome sequencing (CES; n = 146). Variants were interpreted according to the American College of Medical Genetics and Genomics (ACMG) guidelines. Clinical data, including seizure semiology, neuroimaging, EEG findings, and treatment response, were retrospectively reviewed.

Results

A definitive molecular diagnosis (pathogenic or likely pathogenic variants) was established in 89 patients (89/250, 35.6%). Including variants of uncertain significance (VUS) with strong phenotypic correlation raised the diagnostic yield to 57.6% (144/250). The genetic landscape was highly heterogeneous, involving 59 distinct genes. SCN1A was the most frequently implicated gene (n = 11, 12.4% of diagnosed cases), followed by MECP2 and PRRT2 (n = 5 each). Notably, the recurrent PRRT2 c.649dup (p.Arg217Profs*8) frameshift variant was identified in four unrelated Turkish families. Inheritance patterns included autosomal dominant (56/89, 63%), autosomal recessive (20/89, 22%), and X‐linked (13/89, 15%). Diagnostic yields were comparable between WES (38/104, 36.5%) and CES (51/146, 35%) (p = 0.97). Neonatal‐onset epilepsy (p = 0.03) and the presence of autistic features (p = 0.04) were significantly associated with a positive genetic diagnosis.

Conclusion

Our study confirms NGS as a first‐tier diagnostic tool in pediatric epilepsy, revealing a distinctive genetic architecture enriched with novel and population‐specific variants. These findings emphasize (1) a high diagnostic yield (∼36%) supporting first‐tier use of NGS; (2) a distinctive genetic architecture characterized by recurrent PRRT2 variants and an elevated burden of autosomal recessive disorders, likely reflecting regional consanguinity patterns; and (3) a phenotype‐driven framework for prioritizing VUS re‐evaluation in clinical neurology practice. Such data from underrepresented populations are crucial for global genomic databases and directly inform precision medicine and genetic counseling.

Key points.

• NGS provides 35.6% diagnostic yield in unexplained pediatric epilepsy.

• Recurrent PRRT2 c.649dup may reflect either a known mutational hotspot or population‐specific enrichment.

• 22% of diagnoses were autosomal recessive, linked to consanguinity.

• Neonatal onset and autistic features predict higher diagnostic yield.

• Phenotype‐driven VUS reclassification improves clinical utility.

1. INTRODUCTION

Epilepsy affects 0.5%–1% of children worldwide and represents a clinically and genetically heterogeneous group of neurological disorders characterized by recurrent, unprovoked seizures. ¹ , ² Despite advances in neuroimaging and electroencephalography (EEG), a substantial proportion of pediatric epilepsy cases remain etiologically undiagnosed after standard clinical workup. The International League Against Epilepsy (ILAE) now emphasizes the integration of etiology into epilepsy classification, as identifying an underlying cause directly informs treatment selection, prognosis, and genetic counseling. ³ , ⁴

Over the past decade, next‐generation sequencing (NGS), particularly whole‐exome sequencing (WES) and targeted clinical exome sequencing (CES), has transformed the diagnostic pathway for neurogenetic disorders. ⁵ , ⁶ In pediatric epilepsy, NGS has yielded molecular diagnoses in 20%–50% of previously unexplained cases, with the highest yields observed in early‐onset, drug‐resistant, or syndromic presentations. ⁷ , ⁸ These diagnoses not only end the diagnostic odyssey for families but also enable precision therapeutics, such as avoiding sodium channel blockers in SCN1A‐related Dravet syndrome or initiating ketogenic diet in SLC2A1‐associated glucose transporter deficiency. ⁹

However, most large‐scale genomic epilepsy studies have been conducted in populations of European or East Asian ancestry, potentially limiting the generalizability of findings to other ethnic groups. ¹⁰ Emerging evidence indicates that the genetic architecture of epilepsy may differ across ancestries due to population‐specific allele frequencies, consanguinity patterns, and founder mutations. ¹¹ In Turkiye, where consanguineous marriages are relatively common, autosomal recessive disorders are more prevalent, and recurrent pathogenic variants may reflect local founder effects. ¹²

In our current study, WES and CES were performed on a large cohort of 250 pediatric patients with varying ages of onset, diagnosed with epilepsy, and of unknown etiology (metabolic, immune and infectious related etiologies were ruled out). The primary aim was to determine the diagnostic efficiency of NGS in this Turkish pediatric cohort and to characterize the genetic spectrum, including recurrent variants and causative genes, in various epilepsy subgroups. In addition, a variant classification approach based on the American College of Medical Genetics (ACMG) guidelines has enabled the identification of causal variants such as pathogenic/possibly pathogenic (P/LP) and Variant of Uncertain Significance (VUS) in epilepsy‐related genes.

2. MATERIALS AND METHODS

2.1. Study design

This retrospective cohort study included 250 pediatric patients (aged ≤18 years) with epilepsy of unknown etiology (Patients with structural, infectious, immune‐mediated, or metabolic causes identified through neuroimaging, laboratory testing, or clinical history were excluded) referred from the Department of Pediatric Neurology to the Medical Genetics Outpatient Clinic at Pamukkale University Faculty of Medicine between January 2021 and December 2024. These patients were reclassified according to the ILAE updated 2025 classification by retrospectively re‐evaluating the records.

Inclusion criteria were: (1) The patient had a diagnosis of epilepsy. The diagnosis of epilepsy was supported by a detailed seizure history, ictal/interictal EEG recordings, and brain MRI. (2) Seizure onset before 18 years of age. (3) Normal basal metabolic screening (Tandem MS, amino acids, organic acids).

Exclusion criteria included: (1) Patients initially referred with a diagnosis of epilepsy but subsequently determined to have nonepileptic events based on detailed clinical history and follow‐up, such as febrile seizures or breath‐holding spells; (2) Written informed consent was obtained from legal guardians for genetic testing and the use of anonymized clinical and genomic data for research purposes.

2.2. Clinical phenotyping

Comprehensive phenotypic data were extracted from electronic medical records by two experienced pediatric neurologists (blinded to genetic results). Collected variables included:

• Age at seizure onset (categorized as neonatal: 0–28 days; infantile: 1–12 months; toddler: 13–36 months; childhood–juvenile: >36 months)

• Sex

• Family history of epilepsy or neurodevelopmental disorders

• Presence of intellectual disability/developmental delay (ID/DD)

• Behavioral features: autism spectrum disorder (ASD) traits, attention‐deficit/hyperactivity disorder (ADHD) (using DSM‐5 criteria), stereotypies, hyperactivity

• Neurological examination findings: abnormal muscle tone, micro/macrocephaly

• Brain MRI findings (available in 193/250, 77.2%)

• EEG results (available in 197/250, 78.8%): EEG recordings were performed as part of the initial diagnostic evaluation, typically within 6 months of seizure onset. The most representative EEG tracing (demonstrating interictal epileptiform abnormalities or ictal patterns correlating with the predominant seizure type) was selected for analysis in cases with multiple recordings.

• Epilepsy syndrome classification (e.g., developmental and epileptic encephalopathy [DEE])

• Seizure semiology was categorized as focal (F), unknown onset (U), generalized (G), or unclassified according to the updated 2025 ILAE framework, ¹³ based on the most consistent patterns observed during the first year of epilepsy via EEG, caregiver reports, and clinical notes; however, 58 patients (23.2%) were designated as “unclassified” due to insufficient data.

• Seizure semiology describes the ictal event (e.g., generalized tonic–clonic, focal impaired awareness), whereas epilepsy syndrome classification (e.g., Dravet syndrome, developmental and epileptic encephalopathy) integrates seizure types with age of onset, EEG findings, developmental trajectory, and genetic etiology per ILAE criteria.

• Response to anti‐seizure medications (ASM), with drug resistance defined as failure of ≥2 appropriately chosen and dosed ASMs. ¹⁴

2.3. Genomic DNA extraction and sequencing

Genomic DNA was extracted from peripheral blood leukocytes using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). DNA concentration and purity were assessed via spectrophotometry (NanoDrop 2000, Thermo Fisher Scientific).

Next‐generation sequencing was performed using two platforms:

2.3.1. Whole‐exome sequencing (WES) (n = 104)

Library preparation with the Agilent SureSelect Human All Exon V7 kit, followed by paired‐end sequencing on the Illumina NovaSeq 6000 platform (mean coverage depth: ~100×; ≥97% of target bases covered at ≥20×).

2.3.2. Clinical exome sequencing (CES) (n = 146)

Targeted sequencing of a curated panel of >4500 clinically relevant genes using an accredited diagnostic pipeline.

Sequence reads were aligned to the human reference genome (GRCh37/hg19). Variants were called using GATK best practices.

2.4. Bioinformatic analysis and variant interpretation

Variant annotation and filtering were performed using Sophia DDM (v5.10.14) and VarSome (v4.2). Filtering criteria included: Minor allele frequency (MAF) <1% in gnomAD; Predicted deleterious impact (e.g., missense, frameshift, nonsense, splice‐site); Presence in genes with established or definitive disease validity (per ClinGen). All candidate variants were classified according to the American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) guidelines. ¹⁵ Only pathogenic (P) and likely pathogenic (LP) variants in genes with strong genotype–phenotype correlation was considered diagnostic. Variants of uncertain significance (VUS) were recorded but not included in the primary diagnostic yield.

A multidisciplinary team, including clinical geneticists, molecular biologists, and pediatric neurologists, reviewed all P/LP/VUS findings in the context of individual phenotypes.

2.5. Statistical analysis

Categorical variables (e.g., sex, ID/DD, drug resistance, autistic features) were compared between diagnostic subgroups using Fisher's exact test. Odds ratios (OR) with 95% confidence intervals (CI) were calculated where applicable. A two‐sided p‐value <0.05 was considered statistically significant. Analyses were performed using SPSS v28 and R v4.3.

3. RESULTS

3.1. Cohort characteristics and diagnostic yield

Of 250 patients (115 female, 135 male), a definitive molecular diagnosis (P or LP variant) was established in 89 (35.6%). An additional 55 patients (22%) carried VUS, and 106 (42.4%) remained undiagnosed. The diagnostic yield did not differ significantly between WES (38/104, 36.5%) and CES (51/146, 35%) (p = 0.94) (Figure 1).

FIGURE 1.

CES and WES methods diagnostic efficiency.

Age at seizure onset showed marked variation in diagnostic yield: Neonatal: 8/15 (53.3%); infantile: 32/93 (34.4%); toddler: 11/43 (25.6%); childhood–juvenile: 38/99 (38.4%). Neurodevelopmental comorbidities were common: ID/DD: 165/250 (66.0%); autistic features: 10/250 (4.0%); ADHD: 12/250 (4.8%); abnormal muscle tone: 34/250 (13.6%). Microcephaly was observed in 34/250 (13.6%), macrocephaly in 7/250 (2.8%). A family history of epilepsy was present in 56/250 (22.4%). Treatment resistance was present in 28% (70/250) of the cohort, and developmental and epileptic encephalopathy (DEE) was present in 19.6% (49/250). In terms of seizure semiology, unclassified seizures constituted the largest clinical group, reflecting the inherent diagnostic challenges and phenotypic fluidity often encountered in early‐onset epilepsy cases. Generalized tonic–clonic seizures (GTC) were the second most frequent presentation, followed by focal impaired awareness seizures (Table 1).

TABLE 1.

Clinical and demographic characteristics of the study cohort by diagnostic outcome.

	Total number of patients	Genetically diagnosed	VUS	Not diagnosed	p‐Value a
Total number of patients	250	89	55	106
Gender (Female/Male)	115/135	45/44	23/32	47/59
Seizure onset
Neonatal	15	8	1	6	p = 0.03
Infantile	93	32	22	39
Toddler	43	11	13	19
Childhood–Juvenile	99	38	19	42
Family history for epilepsy	56	14	18	24
Intellectual disability/developmental delay (ID/DD)	165	60	35	70	p = 0.50
Head circumference
Microcephaly	34	13	6	15
Macrocephaly	7	4	‐	3
Abnormal Brain MRI b	99/193	32	24	43
Abnormal muscle tone	34	11	11	12
Behavior	42	18	7	17
Autistic behavior	10	7	1	2	p ≈ 0.04
ADHD	12	2	2	8
Stereotypical movements	5	5	‐	‐
Behavioral disorder	13	3	4	6
Hyperactivity	2	1	‐	1
Drug resistance	70	27	14	29	p = 0.46
Type of genetic testing
Whole exome sequencing	104	38	22	44
Clinical exome sequencing	146	51	33	62
Developmental and epileptic encephalopathy (DEE)	49	25	11	13
Classification of epileptic seizures c
Focal (F)
Focal preserved consciousness seizure (FPC)	6	3	‐	3
Focal impaired consciousness seizure (FIC)	42	16	12	14
Focal‐to‐bilateral tonic–clonic seizure (FBTC)	16	4	3	9
Unknown whether focal or generalized (U)
U‐Preserved consciousness seizure (PC)	1	1	‐	‐
U‐ İmpaired consciousness seizure (IC)	7	3	2	2
U‐Bilateral tonic–clonic seizure (BTC)	4	1	1	2
Generalized (G)
Absence seizures (AS)	14	2	6	6
Generalized tonic–clonic seizure (GTC)	61	25	13	23
Absence‐to‐tonic–clonic seizure (ATC)	1	1	‐	‐
Other generalized seizures
Generalized myoclonic seizure (GM)	1	‐	‐	1
Generalized clonic seizure (GC)	5	2	‐	3
Generalized negative myoclonic seizure (GNM)	‐	‐	‐	‐
Generalized epileptic spasms (GES)	11	8	‐	3
Generalized tonic seizure (GT)	13	5	4	4
Generalized atonic seizure (GA)	2	‐	1	1
Generalized myoclonic–atonic seizure (GMA)	8	‐	4	4
Unclassified	58	18	9	31

Abbreviation: ADHD: Attention‐deficit/hyperactivity disorder (DSM‐5 criteria).

^a p‐Value calculated using Fisher's exact test (two‐tailed); statistical significance defined as p < 0.05.

^b MRI data available for 193/250 patients (77.2%).

^c Seizure classification per ILAE 2025 operational criteria. ¹³

3.2. Neuroimaging and neurophysiology

Brain MRI was abnormal in 99/193 (51.3%) of patients with available imaging. Among the genetically diagnosed patients with MRI data, 45.7% (37/81) had structural abnormalities (e.g., atrophy, white matter changes, cortical dysplasia).

EEG was available for 197 patients; epileptiform abnormalities were observed in 117 (59.4%). Of these, 53 (53/117, 45.3%) received a definitive genetic diagnosis, while 36 (45%) of the 80 patients with normal EEG received a genetic diagnosis, which was not statistically significant (p = 0.09).

3.3. Genetic findings and variant spectrum

In our cohort, a definitive molecular diagnosis was made in 89 (89/250, 35.6%) patients. The molecular profile of these patients included 83 different variants (LP/P) in 59 different genes, and the allelic variant types were as follows: Missense: 38 (42%); Frameshift: 27 (30%); Nonsense: 19 (21%); Splice site: 5 (5%); Synonym: 2 (2%) (Figure 2A). Thirty‐two of the detected variants were novel and had not been previously reported. The inheritance pattern of the patients was autosomal dominant (AD) in 63% (56/89), autosomal recessive (AR) in 22% (20/89), and X‐linked (XL) in 15% (13/89) (Figure 2B). The SCN1A gene, detected in 11 cases, represented the largest proportion among the genes with the most frequently identified variants. The majority of SCN1A variants were SNVs, often missense variants. Other frequently recurring genes included PRRT2 (n = 5). Specifically, the PRRT2 frameshift variant c.649dup (p.Arg217Profs*8) was found in four unrelated patients in this cohort, indicating a potentially recurring variant. Other genes, in order of frequency, were KCNQ2 (n = 3), SCN2A (n = 2), NSD1 (n = 2), SLC2A1 (n = 2), and ZEB2 (n = 2). In the AR inheritance model, AP4M1 (n = 3), TBC1D23 (n = 2), and WWOX (n = 2) were the most frequently encountered genes. Among the 9 genes showing X‐linked inheritance, the most common were MECP2 (n = 5) and PCDH19 (n = 2).

FIGURE 2.

In cases with definitive molecular diagnosis. (A) Types of variants identified. (B) Observed inheritance patterns. (C) Distribution of identified genes according to pathway information.

Furthermore, 53 promising VUS variants were identified in 51 different genes in 55 of our patients. Twenty‐six of these variants had not been previously reported. Although there was no evidence to elevate the classification of these rare VUS variants according to ACMG criteria, they were considered because they affect a conserved amino acid, may be deleterious, and may be associated with a range of diseases matching the patients' phenotype. Table 2 provides a comprehensive list of all identified genes and variants (LP/P/VUS). To elucidate the genetic pathogenesis of epilepsy, genes with variants (P/LP) identified in our cohort were divided into 7 different groups according to their function: cell signaling, growth, and mTOR/RAS–MAPK pathways; ion transport and membrane excitability; metabolism, mitochondrial/lysosomal function, and DNA repair; protein homeostasis and extracellular matrix; regulation of gene transcription, chromatin, and RNA processing; regulation of synaptic processes; and structural, cytoskeleton, axonal transport, adhesion, and myelin. The largest segment consisted of ion transport and membrane excitability genes (SCN1A, SCN2A, KCNQ2, CACNA1A, GRIN2A, GRIN1, GRIA2, GABRB3, GRIN2B, GRIN2D). The second largest segment was regulated by genes involved in transcription regulation, chromatin processing, and RNA processing (MECP2, NSD1, KMT2D, ZEB2, PUF60, CHD3, SMC1A, ZNF292, SON, TSEN54, FOXG1, SMARCA2, MYT1L, ATRX, ASH1L), while protein homeostasis and extracellular matrix genes (USP9X, MGP) were present in only two patients and had the lowest presence (Figure 2C). The gene group distribution in VUS was similar to LP/P variants, with the highest proportion found in ion transport and membrane excitability genes.

TABLE 2.

Pathogenic, likely pathogenic, and select variants of uncertain significance identified in the cohort.

Case ID	Gender	Gene	Variant type	Nucleotide change	Predicted effect	Zyg	Novel/reported	ACMG criteria applied	ACMG Clas.	Seizure onset	Classification of epileptic seizures a	RTM
45	F	SCN1A	Missense	c.4222T>C	p.Trp1408Arg	Het	Novel	PM1, PP2, PP3, PM2	LP	Infantile	GTC
64	M	SCN1A	Synonymous	c.1170G>A	p.Leu390=	Het	Reported ClinVar ID:946265	PS2, PM2, PP3	LP	Infantile	GT	R
85	F	SCN1A	Nonsense	c.1624C>T	p.Arg542*	Het	Reported ClinVar ID:189848	PVS1, PM2, PS2	P	Childhood	GTC
104	F	SCN1A	Frameshift	c.1496del	p. Asn499Ilefs*45	Het	Novel	PVS1, PM2	LP	Childhood	FBTC	R
127	M	SCN1A	Nonsense	c.1702C>T	p.Arg568*	Het	Reported ClinVar ID:265254	PVS1, PM2, PS2	P	Infantile	Unclassified
136	F	SCN1A	Missense	c.82C>T	p.Arg28Cys	Het	Reported ClinVar ID:206802	PS4, PM2, PM5, PP3PP2	P	Childhood	GTC
165	F	SCN1A	Missense	c.3709T>C	p.Phe1237Leu	Het	Novel	PM1, PP2, PM5, PP3	LP	Childhood	FIC
180	M	SCN1A	Missense	c.5132G>A	p.Gly1711Asp	Het	Novel	PM1, PP2, PM2, PM5, PP3	LP	Toddler	Unclassified	R
187	F	SCN1A	Frameshift	c.1655del	p.Pro552fs*6	Het	Novel	PVS1, PM2	LP	Infantile	GC
206	M	SCN1A	Missense	c.4015G>T	p.Ala1339Ser	Het	Novel	PM1, PP2, PM2, PM5, PP3	LP	Childhood	PC
238	M	SCN1A	Missense	c.2051C>T	p.Thr684Ile	Het	Reported ClinVar ID:643782	PM2, PP3, PP2	VUS	Infantile	GTC
244	M	SCN1A	Frameshift	c.765_766del	p.Phe256Leufs*20	Het	Reported ClinVar ID:280823	PVS1, PS4, PM2	P	Infantile	Unclassified	R
47	F	MECP2	Missense	c.433C>T	p.Arg145Cys	Het	Reported ClinVar ID:11809	PS4, PS2, PM2, PM5, PM1, PP3	P	Toddler	Unclassified	R
101	F	MECP2	Missense	c.437C>G	p.Ser146Cys	Het	Reported ClinVar ID:143562	PS3, PS2, PM2, PM5, PM1, PP3	P	Childhood	FBTC
128	F	MECP2	Nonsense	c.916C>T	p.Arg306*	Het	Reported ClinVar ID:11819	PVS1, PS2, PS3, PS4, PM2	P	Childhood	GTC	R
140	F	MECP2	Missense	c.509C>G	p.Thr170Arg	Het	Reported ClinVar ID:2135175	PS4, PM2, PM5, PM1, PP3	LP	Infantile	GTC
245	F	MECP2	Nonsense	c.1200_1243del	p.Pro401*	Het	Reported ClinVar ID:143406	PVS1, PP1, PS2, PM2	P	Childhood	Unclassified
58	F	PRRT2	Frameshift	c.649dup	p.Arg217Profs*8	Het	Reported ClinVar ID:65758	PVS1, PS3, PS4, PP1, PM2	P	Infantile	GTC
99	F	PRRT2	Frameshift	c.649dup	p.Arg217Profs*8	Het	Reported ClinVar ID:65758	PSV1, PS4, PS3, PP1, PM2	P	Infantile	GTC
100	F	PRRT2	Frameshift	c.649dup	p.Arg217Profs*8	Het	Reported ClinVar ID:65758	PVS1, PS4, PS3, PP1, PM2	P	Childhood	Unclassified
197	M	PRRT2	Synonymous	c.1011C>A	p.Gly337=	Het	Reported *ClinVar ID:372477	PS2, PS4, BS1	P	Infantile	Unclassified
247	M	PRRT2	Frameshift	c.649dup	p.Arg217Profs*8	Het	Reported ClinVar ID:65758	PVS1, PS3, PS4, PP1, PM2	P	Infantile	IC
26	F	PCDH19	Missense	c.50C>T	p.Thr17Met	Het	Reported ClinVar ID:2434604	PM1, PP2, PM2	VUS	Toddler	GT
89	M	PCDH19	Missense	c.1238G>T	p.Arg413Leu	Hemi	Novel	PM1, PP2, PM2, PP3	LP	Childhood	GT	R
130	F	PCDH19	Nonsense	c.1198G>T	p.Glu400*	Het	Novel	PVS1, PM2	LP	Infantile	GTC	R
65	M	AP4M1	Nonsense	c.1264C>T	p.Arg422*	Hom	Reported dbSNP ID: rs1440310254	PVS1, PM2	LP	Neonatal	FIC
66	F	AP4M1	Nonsense	c.1264C>T	p.Arg422*	Hom	Reported dbSNP ID: rs1440310254	PVS1, PM2	LP	Childhood	Unclassified
69	M	AP4M1	Nonsense	c.1012C>T	p.Arg338*	Hom	Reported ClinVar ID:209980	PVS1, PM3, PM2	P	Infantile	GTC
71	M	SCN2A	Missense	c.4954G>A	p.Ala1652Thr	Het	Novel	PM1, PP2, PM2, PM5, PP3	LP	Childhood	Unclassified
80	M	SCN2A	Missense	c.4418T>C	p.Ile1473Thr	Het	Reported ClinVar ID:427011	PS4, PM1, PP2 PM2, PM5, PP3	P	Childhood	GTC
250	M	SCN2A	Missense	c.2300T>C	p.Ile767Thr	Het	Reported ClinVar ID:1318991	PM2, PP3, PP2	VUS	Infantile	Unclassified
119	F	KCNQ2	Nonsense	c.2425C>T	p.Gln809*	Het	Reported ClinVar ID:813752	PVS1, PS4, PM2	P	Childhood	FBTC	R
152	M	KCNQ2	Missense	c.629G>C	p.Arg210Pro	Het	Reported ClinVar ID:280614	PS4, PM1, PP2, PM2, PM5, PP3	P	Neonatal	GTC
219	M	KCNQ2	Missense	c.1037T>C	p.Phe346Ser	Het	Novel	PM1, PP2, PM2, PP3	LP	Neonatal	GC
133	M	CNTNAP2	Missense	c.963C>G	p.Phe321Leu	Hom	Reported ClinVar ID:546436	PM2	VUS	Infantile	FIC
134	M	CNTNAP2	Missense	c.963C>G	p.Phe321Leu	Hom	Reported ClinVar ID:546436	PM2	VUS	Infantile	GTC
182	F	CNTNAP2	Missense	c.1560C>G	p.Cys520Trp	Hom	Novel	PM2, PP3	VUS	Childhood	Unclassified
3	M	TBC1D23	Frameshift	c.743_744del	p.Gln248Argfs*43	Hom	Reported ClinVar ID:4293118	PVS1, PM3, PM2	P	Infantile	GTC
4	M	TBC1D23	Frameshift	c.743_744del	p.Gln248Argfs*43	Hom	Reported ClinVar ID:4293118	PVS1, PM3, PM2	P	Toddler	FIC
5	F	TBCD	Frameshift/ Missense	c.1575del/ c.2848C>T	p.His526Metfs*6/ p.Pro950Ser	Com.Het	Novel/ Novel	PVS1, PM2/PM2	LP/ VUS	Toddler	GTC	R
239	F	TBCD	Frameshift/ Missense	c.1575del/ c.2848C>T	p.H526Mfs*6/ p.P950S	Com.het	Novel/ Novel	PVS1, PM2/ PM2	LP/ VUS	Childhood	GT	R
9	M	NSD1	Frameshift	c.4858_4859del	p.Val1620ProfsTer27	Het	Novel	PVS1, PM2	LP	Infantile	GTC
137	M	NSD1	Splice donor	c.4378+3_4378+6del	Splicing	Het	Reported ClinVar ID:565696	PS2, PP3, PM2	LP	Childhood	Unclassified
10	M	CACNA1A	Missense	c.6743C>T	p.Ser2254Leu	Het	Reported ClinVar ID:1514365	PM2, PP3, PP2	VUS	Toddler	GTC
20	F	CACNA1A	Frameshift	c.2045dup	p.Val683Argfs*98	Het	Reported ClinVar ID:2021246	PVS1, PM2, PP5	P	Neonatal	FPC
13	M	DYNC1H1	Missense	c.3345G>C	p.Lys1115Asn	Hom	Novel	PM2, PP2	VUS	Neonatal	GT	R
98	F	DYNC1H1	Frameshift	c.947_948insGACAC	p.Phe316Leufs*9	Het	Novel	PVS1, PM2	LP	Infantile	GTC	R
15	F	GRIN2A	Missense	c.2327A>G	p.Asp776Gly	Het	Novel	PM2, PM5, PP3, PM1	LP	Childhood	BTC	R
179	F	GRIN2A	Missense	c.1985T>A	p.Val662Glu	Het	Novel	PM2, PM1, PP3	VUS	Toddler	AS	R
29	F	WWOX	Nonsense	c.790C>T	p.Arg264*	Hom	Reported ClinVar ID:241105	PVS1, PM3, PM2	P	Infantile	GES	R
51	M	WWOX	Nonsense	c.790C>T	p.Arg264*	Hom	Reported ClinVar ID:241105	PVS1, PM3, PM2	P	Neonatal	GES
78	M	GRIN1	Missense	c.1913G>C	p.Gly659Ala	Het	Reported ClinVar ID:975942	PS2, PM1, PP2, PM2	LP	Infantile	GES	R
167	F	GRIN1	Missense	c.425C>T	p.Pro142Leu	Het	Novel	PM2, PP3, PP2	VUS	Infantile	Unclassified
114	F	SLC2A1	Missense	c.697G>C	p.Gly233Arg	Het	Reported ClinVar ID:471348	PM1, PP2, PM2, PP3	LP	Infantile	Unclassified
186	M	SLC2A1	Missense	c.695G>A	p.Arg232His	Het	Reported ClinVar ID:623683	PS4, PM1, PP2, PM2, PM5, PP3	P	Childhood	Unclassified
131	F	KMT2D	Frameshift	c.2029del	p.Glu677Serfs*253	Het	Novel	PVS1, PM2	LP	Childhood	Unclassified
144	F	KMT2D	Missense	c.8642G>A	p.Arg2881Gln	Het	Reported ClinVar ID:2891063	PM2, PP2, BP6	VUS	Toddler	GTC
183	F	GAD1	Splice‐altering	c.1120‐493T>A	splicing	Hom	Novel	PM2	VUS	Infantile	GT
184	F	GAD1	Splice‐altering	c.1120‐493T>A	splicing	Hom	Novel	PM2	VUS	Infantile	GT
202	F	RORB	Missense	c.334A>T	p.Ser112Cys	Het	Reported ClinVar ID:1997337	PM2	VUS	Childhood	GMA	R
222	F	RORB	Missense	c.835C>T	p.Arg279Trp	Het	Reported ClinVar ID:1013334	PM2, PP3	VUS	Toddler	AS
18	M	ZEB2	Missense	c.3133C>A	p.His1045Asn	Het	Novel	PM1, PP2, PM2 PM5, PP3	LP	Childhood	FIC
223	F	ZEB2	Missense	c.898C>T	p.His300Tyr	Het	Reported ClinVar ID:3984323	PM2, PM5, PP3 PP2	LP	Childhood	FIC
2	F	GALC	Missense	c.857G>A	p.Gly286Asp	Hom	Reported ClinVar ID:30619	PM3, PS3, PM1, PM2, PM5, PP3	P	Neonatal	GTC	R
6	M	WDR81	Missense	c.2206C>T	p.Leu1939Phe	Hom	Novel	PM2	VUS	Infantile	GTC	R
7	M	SACS	Missense	c.829A>G	p.Met424Val	Hom	Reported dbSNP ID: rs1286940495	PM2	VUS	Infantile	GTC
12	M	GLB1	Missense	c.110G>A	p.Ser37Asn	Hom	Novel	PM1, PM2	VUS	Infantile	FIC
14	F	PTPN11	Missense	c.164A>G	p.Lys55Arg	Het	Reported ClinVar ID:1397955	PM1, PP2, PM2	VUS	Childhood	Unclassified
16	F	GRIA2	Missense	c.1939G>C	p.Val647Leu	Het	Reported ClinVar ID:929842	PS2, PM1, PP2, PM2, PM5	P	Neonatal	FIC	R
19	M	IL1RAPL	Nonsense	c.148C>T	p.Arg50*	Hemi	Novel	PVS1, PM3, PM2	P	Childhood	FPC
21	F	RELN	Missense	c.6976G>A	p.Asp2326Asn	Het	Reported ClinVar ID:651420	PM2, BP6	VUS	Childhood	IC
23	F	PUF60	Frameshift	c.1263del	p.Glu422Argfs*13	Het	Novel	PVS1, PM2	LP	Childhood	FBTC
24	F	CACNA1B	In‐frame Deletion	c.5823_5831del	p.Arg1943_Gln1945del	Het	Novel	PM2, PM4	VUS	Infantile	FIC	R
27	F	ITPA	Splice donor	c.‐214+1G>A	splicing	Het	Reported ClinVar ID:646228	PVS1, PM3, PM2	P	Toddler	GTC
28	M	MTHFS	Missense	c.157T>C	p.Ser53Pro	Hom	Novel	PM2	VUS	Childhood	FBTC	R
31	M	SCN5A	Missense	c.4009G>A	p.Val1337Ile	Het	Reported ClinVar ID:918378	PM2, PM1, PP3	VUS	Infantile	FIC
32	M	CHD3	Nonsense	c.5099C>G	p.Ser1641*	Het	Novel	PVS1, PM2	LP	Infantile	FIC
34	F	GALNT2	Missense	c.1084C>T	p.Arg362Trp	Hom	Reported dbSNP ID: rs774689179	PM2, PP3	VUS	Childhood	IC
35	M	KCNQ1	İntronic	c.1033‐85A>G	‐	Het	Novel	PM2	VUS	Toddler	Unclassified
36	M	ZDHHC9	Nonsense	c.892C>T	p.Arg298*	Hemi	Reported ClinVar ID:2228665	PVS1, PM3, PM2	P	Childhood	IC
37	F	PLA2G6	Missense	c.2251G>A	p.Glu751Lys	Hom	Reported ClinVar ID:2138447	PM3, PM1, PM2	LP	Toddler	GT
38	M	KIF1A	Missense	c.296C>T	p.Thr99Met	Het	Reported ClinVar ID:30169	PS2, PP2, PM2 PM5, PP3	P	Childhood	Unclassified
39	F	USP9X	Splice site	Intron 24 c.3684+2T>G	splicing	Het	Novel	PVS1, PM2	LP	Infantile	GTC	R
40	M	GABRB3	Missense	c.867C>G	p.Ile289Met	Het	Novel	PP2, PM2, PP3	LP	Infantile	FIC	R
44	M	CHRNA4	Missense	c.1453C>T	p.Arg485Trp	Het	Reported ClinVar ID:1003101	PM2	VUS	Infantile	GMA	R
48	F	SMC1A	Missense	c.1193G>A	p.Arg398Gln	Het	Reported ClinVar ID:159938	PS4, PM2, PM5, PP3, PP2	LP	Toddler	GTC
49	M	KATNB1	Missense	c.1835G>A	p.Arg612Lys	Hom	Novel	PM2, PP3	VUS	Childhood	FIC
52	F	ZNF292	Frameshift	c.6305_6306del	p.Leu2102Argfs*53	Het	Novel	PVS1, PM2	LP	Infantile	GES
57	M	FGFR2	Missense	c.532C>T	p.Arg178Cys	Het	Reported ClinVar ID:2226118	PM2, PP3, PP2	VUS	Infantile	FIC
61	M	RHOBTB2	Deletion	c.2059_2061del	p.Lys709del	Het	Novel	PM2, PM4	VUS	Childhood	GA	R
67	F	PDHA1	Missense	c.788G>C	p.Arg301Pro	Het	Reported ClinVar ID:1184474	PM2, PM5, PP3, PP2	LP	Infantile	GTC	R
68	F	IGHMBP2	Missense	c.1540G>A	p.Glu514Lys	Hom	Reported ClinVar ID:9112	PM3, PM2, PP3	P	Infantile	FIC
75	F	CACNA2D2	Missense	c.1046A>T	p.Asn349Ile	Hom	Novel	PM2, PP3	VUS	Infantile	FIC
76	M	KIF5C	Missense	c.929A>G	p.Asn310Ser	Het	Novel	PM2, PP3, PP2	VUS	Childhood	GTC
77	F	CKAP2L	Frameshift	c.981_982del	p.Arg327Serfs*16	Hom	Novel	PVS1, PM2	LP	Childhood	Unclassified
79	F	PNKP	Frameshift	c.1253_1269dup	p.Thr424Glyfs*49	Hom	Reported ClinVar ID: 4847	PVS1, PM3, PP1, PS3, PM2	P	Childhood	GT
81	F	NPRL3	Frameshift	c.511del	p.Leu171Trpfs*46	Het	Novel	PVS1, PM2	LP	Infantile	GES	R
83	M	MGP	Nonsense	c.43del	p.Val15*	Hom	Reported ClinVar ID:14341	PVS1, PM2, PP5	P	Infantile	FIC
84	M	PIK3R2	Missense	c.1117G>C	p.Gly373Arg	Het	Reported *ClinVar ID:39808	PS1, PM2	LP	Infantile	GES
86	M	DDX3X	Missense	c.1272A>C	p.Glu424Asp	Hemi	Novel	PM1, PM2	VUS	Childhood	GTC
87	M	GRIN2B	Missense	c.2453T>C	p.Met818Thr	Het	Reported ClinVar ID:245960	PS2, PS3, PP2, PM2, PM5, PP3	P	Infantile	GES
88	F	PTEN	Missense	c.611C>G	p.Pro204Arg	Het	Novel	PM1, PP2, PM2, PM5, PP3	LP	Childhood	IC
92	M	SON	Frameshift	c.5753_5756del	p.Val1918Glufs*87	Het	Reported ClinVar ID:252929	PVS1, PS2, PM2	P	Infantile	FIC
94	M	YEATS2	Missense	c.2738C>T	p.Ala913Val	Het	Reported dbSNP ID: rs769948328	PM2, PP3	VUS	Childhood	AS
95	M	RNASEH2A	Missense	c.872G>A	p.Arg291His	Hom	Reported ClinVar ID:126401	PM2	VUS	Childhood	FIC
96	M	TRIO	Missense	c.8897G>A	p.Gly2966Glu	Het	Reported dbSNP ID: rs903212005	PM2, PP2	VUS	Toddler	FBTC
97	M	HCFC1	Missense	c.1024G>A	p.Gly342Ser	Hem	Novel	PM2, PP2	VUS	Childhood	Unclassified
102	F	ACOX1	Nonsense	c.768T>A	p.Tyr256*	Het	Novel	PVS1, PM2	LP	Neonatal	FIC
103	M	TSEN54	Missense	c.919G>T	p.Ala307Ser	Hom	Reported ClinVar ID:2120	PM3, PP1, PM2	P	Toddler	GTC	R
106	F	HGSNAT	Missense	c.1277G>A	p.Gly426Glu	Hom	Novel	PM2, PP3	VUS	Childhood	FIC	R
107	M	FOXG1	Frameshift	c.460dup	p.Glu154Glyfs*301	Het	Reported ClinVar ID:95268	PVS1, PS4, PM2	P	Childhood	FIC	R
109	M	MICAL1	Nonsense	c.2910G>A	p.Trp970*	Het	Reported ClinVar ID:2030661	PVS1, PM2	LP	Childhood	AS
110	M	SMARCA2	Missense	c.3449C>T	p.Pro1150Leu	Het	Novel	PP2, PM2, PP3	LP	Infantile	GTC
112	M	SLC6A1	Missense	c.863C>T	p.Ala288Val	Het	Reported ClinVar ID:192372	PS3, PS2, PP2, PM2, PM5, PP3, PP5	P	Childhood	FIC
113	F	NHLRC1	Missense	c.436G>A	p.Asp146Asn	Hom	Reported ClinVar ID:162618	PM3, PP1, PS3, PM2, PM5, PM1	P	Childhood	GTC
115	M	KCNQ5	Missense	c.1810G>A	p.Gly623Ser	Het	Reported *ClinVar ID:1390052	PM2, PP3, PP2	VUS	Toddler	GTC
121	M	SCN11A	Nonsense	c.3448C>T	p.Arg1150*	Het	Reported ClinVar ID:967210	PM2	VUS	Toddler	AS	R
126	F	DOCK7	Missense	c.2195C>G	p.Thr732Arg	Hom	Reported ClinVar ID:1916429	PM2, PP2	VUS	Infantile	FIC	R
135	F	NEXMIF	Frameshift	c.1352_1359del	p.Asp451GlyfsTer2	Het	Novel	PVS1, PM2	LP	Toddler	ATC	R
138	M	KDM5C	Missense	c.56C>A	p.Pro19His	Hemi	Novel	PM2, PP3, PP2	VUS	Toddler	GTC
141	M	SCN9A	Missense	c.2734G>A	p.Asp912Asn	Het	Reported ClinVar ID:1387587	PM2, PP3	VUS	Childhood	AS
142	F	MYT1L	Frameshift	c.812dup	p.His272Thrfs*11	Het	Novel	PVS1, PM2	LP	Childhood	GTC
145	M	SYN1	Missense	c.1966C>A	p.Pro656Thr	Hemi	Novel	PM2, BP4	VUS	Childhood	AS
153	M	SCN1B	Missense	c.538C>T	p.Leu180Phe	Het	Reported ClinVar ID:1371605	PM2	VUS	Childhood	Unclassified	R
156	M	ATRX	Splice site	c.7201‐2A>G	Splicing	Hem	Reported ClinVar ID:11737	PVS1, PM2, PP5	LP	Childhood	Unclassified
162	F	AMPD2	Missense	c.813G>A	p.Met271Ile	Hom	Novel	PM2, PP2	VUS	Toddler	Unclassified
169	F	SYNGAP1	Missense	c.3677A>G	p.Gln1226Arg	Het	Novel	PM2, PP2	VUS	Infantile	GTC
171	M	APTX	Frameshift	c.940_956del	p.Lys314Serfs*2	Hom	Reported ClinVar ID:870836	PVS1, PM2, PP5	LP	Childhood	Unclassified
172	M	SAMD9	Nonsense	c.4299del	p.Leu1434*	Het	Reported dbSNP ID: rs1365970571	PVS1, PM2	LP	Toddler	FIC
178	M	CLPB	Missense	c.422C>A	p.Ala141Asp	Hom	Novel	PM2	VUS	Infantile	GTC
181	M	KCNT1	Missense	c.274G>A	p.Val92Ile	Het	Reported ClinVar ID:804967	PM2	VUS	Infantile	GMA	R
188	M	PIGG	Nonsense	c.1897C>T	p.Arg633*	Hom	Reported ClinVar ID:2871376	PVS1, PM2, PP5	P	Toddler	Unclassified
191	F	KAT8	Missense	c.3G>C	p.Met1Ile	Het	Novel	PM2	VUS	Toddler	GMA	R
195	F	KCNJ10	Missense	c.896C>A	p.Ser299Tyr	Hom	Novel	PM2, PP3	VUS	Infantile	GTC
196	F	CLCN4	Missense	c.1673C>G	p.Thr558Ser	Het	Novel	PM2, PM1, PP2, PP3	VUS	Toddler	FBTC
199	M	ASH1L	Frameshift	c.914_915del	p.V305Efs*13	Het	Novel	PVS1, PM2	LP	Childhood	AS
200	M	MRAP	splice donor	c.106+1delG	splicing	Hom	Reported ClinVar ID:444068	PVS1, PM3, PM2	P	Childhood	FIC
207	M	CPA6	Missense	c.799G>A/ c.619C>G	p.Gly267Arg/ p.Gln207Glu	Com.Het	Reported	PM2, BS2, PP3, BP6/PM2, BS2, BP6	VUS/VUS	Infantile	BTC
217	F	GRIN2D	Frameshift	c.120del	p.Gly42Alafs*41	Het	Novel	PVS1, PM2	LP	Infantile	GES	R
224	F	SETD1A	Missense	c.1010C>T	p.Ser337Phe	Het	Reported dbSNP ID: rs776546240	PM2	VUS	Childhood	FIC
225	M	STXBP1	Missense	c.1652G>A	p.Arg551His	Het	Reported ClinVar ID:566474	PS2, PP2, PM2, PM5, PP3	P	Infantile	GTC	R
230	M	MPZ	Missense	c.106A>T	p.Arg36Trp	Het	Reported ClinVar ID:221065	PS4, PP1, PS3, PM1, PM2, PM5	P	Childhood	FPC	R
231	F	MACF1	Missense	c.21454C>T	p.Arg7152Trp	Het	Reported dbSNP ID: rs768877074	PM2, PP3, PP2	VUS	Infantile	FIC
248	M	CLCN2	Missense	c.703C>T	p.Arg235Trp	Het	Reported dbSNP ID: rs199782817	PM2, PP3, PM5	VUS	Childhood	Unclassified

Note: Full table includes all 98 P/LP and 47 clinically relevant VUS cases; only representative entries shown.

Abbreviations: AD, autosomal dominant; AR, autosomal recessive; ATC, absence‐to‐tonic–clonic; BTC, bilateral tonic–clonic; Com.Het, compound heterozygous; FBTC, focal‐to‐bilateral tonic–clonic; FIC, focal impaired consciousness; GC, generalized clonic; GES, generalized epileptic spasms; GMA, generalized myoclonic–atonic; GT, generalized tonic; GTC, generalized tonic–clonic; Hemi, hemizygous; Het, heterozygous; Hom, homozygous; IC, impaired consciousness; LP, likely pathogenic; P, pathogenic; PC, preserved consciousness; RTM, Response to treatment; VUS, variant of uncertain significance; XL, X‐linked; Zyg, zygosity.

^a Seizure classification follows ILAE 2025 operational criteria. ¹³ Epilepsy syndrome diagnoses (e.g., Dravet syndrome, DEE) were assigned based on comprehensive clinical evaluation including seizure types, EEG, neuroimaging, and developmental history. FLBFR = Drug‐resistant: Failure of ≥2 appropriately chosen and dosed anti‐seizure medications. ¹⁴ Cases marked with ‘R’ include the 12 patients in whom genetic diagnosis directly informed therapeutic decisions (see Section 3.5).

3.4. Clinical determinants of diagnosis

Statistical analysis revealed several clinical features associated with a higher probability of reaching a definitive molecular diagnosis (detection of P/LP variants):

3.4.1. Neonatal seizure onset

Patients with neonatal epilepsy onset showed significantly higher diagnostic yield (8/15, 53.3%) compared to other age groups (range 11/43, 25.6% vs. 38/99, 38.4%), confirming that early onset is a strong predictor of genetic etiology (p = 0.03).

3.4.2. Autistic features

The presence of autistic behaviors was significantly associated with a positive genetic diagnosis (P/LP) (7/10, 70.0% vs. 82/240, 34.1%; OR = 5.6, 95% CI: 1.2–25.7; p ≈ 0.04).

3.4.3. Neurodevelopmental delay (ID/DD)

Although a high proportion of patients with a definitive diagnosis (60/165, 36.4%) had ID/DD, this proportion was not statistically significant when compared to patients without this characteristic (29/85, 34.1%) (p = 0.50).

3.4.4. Drug resistance

No significant difference in diagnostic yield was observed between drug‐resistant (27/70, 38.6%) and non‐drug‐resistant (62/180, 34.4%) patients (p = 0.46).

3.4.5. Neuroimaging and EEG

Although abnormal MRI findings were common among definitively diagnosed cases (32 out of 73 patients with available MRI data had abnormal MRI findings, 43.8%), structural abnormalities were not a statistically significant predictor of genetic diagnosis in this cohort (p > 0.05), possibly due to the inclusion of patients with acquired structural lesions. In the cohort (n = 197) that had access to EEG imaging, 59.4% (117/197) had abnormal EEG findings, and 45.3% (53/117) of these were diagnosed with a definitive molecular diagnosis.

3.5. Clinical utility of genetic diagnosis

In 12 of 89 diagnosed patients (13.5%), genetic results directly informed therapeutic decisions: (1) SCN1A‐positive cases (n = 9) avoided sodium channel blockers (e.g., carbamazepine, lamotrigine) to reduce seizure exacerbation risk; (2) SLC2A1 variants (n = 2) prompted ketogenic diet trials with subsequent seizure reduction; (3) MECP2 diagnosis (n = 1) enabled Rett syndrome‐specific surveillance including scoliosis monitoring and cardiac screening. Of these 12 patients, 6 were drug‐resistant (indicated with ‘R’ in Table 2), highlighting how genetic diagnosis can guide precision therapy in refractory cases.

4. DISCUSSION

This study reports a diagnostic yield of 35.6% (89/250) using next‐generation sequencing (NGS) in a large, unselected Turkish pediatric epilepsy cohort of unknown etiology a rate that aligns with or exceeds recent international benchmarks. ⁷ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ When clinically correlated variants of uncertain significance (VUS) are included, the potential diagnostic yield rises to 57.6% (144/250), directly informing precision therapeutics in 13.5% of diagnosed cases (e.g., sodium channel blocker avoidance in SCN1A, ketogenic diet initiation in SLC2A1). Beyond confirming NGS as a first‐tier diagnostic modality, our findings offer critical insights into genetic architecture, population‐specific variants, and clinical‐actionable implications in early‐onset epilepsy.

4.1. Genetic landscape and population‐specific architecture

The genetic spectrum in our cohort was remarkably heterogeneous, with 59 different genes and 83 variants (136 different variants when VUSs are included) playing a role in definitive diagnoses. This mirrors the known complexity of pediatric epilepsy but also reveals distinct population‐level patterns. Notably, the recurrent identification of the PRRT2 c.649dup variant in four unrelated families may reflect either a known mutational hotspot or a potential population‐specific enrichment in the Turkish cohort. The hypothesis that would require haplotype analysis and comparison with population frequency databases for confirmation. Infantile onset seizures were observed in 3 of the 4 patients; this frameshift variant is a well‐established hotspot for benign familial infantile epilepsy (BFIE) and paroxysmal kinesigenic dyskinesia (PKD), ¹⁸ , ²¹ yet its recurrence in our cohort may reflect regional haplotype conservation, possibly driven by high consanguinity rates (prevalence ~20% in Turkiye). This observation underscores the need for population‐tailored variant databases and highlights how underrepresented cohorts can reveal geographically enriched pathogenic alleles.

Given the elevated burden of autosomal recessive disorders (22% of diagnoses) in our cohort likely reflecting regional consanguinity patterns we recommend that NGS panels deployed in populations with high consanguinity rates should prioritize comprehensive coverage of AR epilepsy genes, particularly those associated with developmental and epileptic encephalopathy (e.g., AP4M1, TBC1D23, WWOX). ² , ²²

While some pediatric epilepsy cohort studies report that women may be more likely to receive a diagnosis than men, almost twice as often ²³ , ²⁴ ; there are also large cohort studies reporting no difference between genders. ¹⁹ , ²⁵ In our patient population, no statistically significant difference was found between female and male patients in terms of diagnosis rates (p > 0.05).

4.2. Genotype–phenotype correlations and DEE spectrum

Forty‐nine patients (49/250, 19.6% of the entire cohort; 25 with P/LP variants, 11 with VUS, and 13 undiagnosed) were clinically classified as having developmental and epileptic encephalopathy (DEE), confirming that DEE constitutes a substantial proportion of severe pediatric epilepsy cases. Notably, one male patient (Case 89) carried a hemizygous PCDH19 missense variant (c.1238G>T, p.Arg413Leu). PCDH19‐related epilepsy typically follows an atypical X‐linked inheritance pattern, wherein heterozygous females are affected while hemizygous males are usually spared due to a “cellular interference” mechanism requiring coexistence of PCDH19‐expressing and nonexpressing neurons. Rare affected males have been reported and may result from: (1) somatic mosaicism confined to brain tissue; (2) skewed X‐chromosome inactivation in carrier mothers with transmission to sons; or (3) dominant‐negative effects of specific missense changes. ²⁶ , ²⁷ Parental samples were unavailable to confirm inheritance or mosaicism; however, this case aligns with prior reports of mosaic or hemizygous PCDH19 males with epilepsy. It is important to distinguish between seizure semiology (the ictal event) and epilepsy syndrome diagnosis (the broader clinical entity). For example, SCN1A variants were associated with diverse seizure semiologies (focal‐to‐bilateral tonic–clonic (fbtc), generalized tonic–clonic (GTC), unclassified) but consistently met criteria for Dravet syndrome or DEE based on early onset, developmental regression, and EEG findings. Similarly, KCNQ2 variants presented with neonatal GTC or Generalized Clonic (GC) seizures, consistent with KCNQ2‐related neonatal epilepsy. This distinction underscores that seizure semiology represents a single dimension of the phenotype, whereas syndrome diagnosis integrates multiple clinical domains over time. The predominance of ion channel genes (SCN1A, SCN2A, KCNQ2, CACNA1A) in this subgroup is consistent with global reports, but our data add granularity: all 11 SCN1A‐positive patients exhibited severe phenotypes, with 9 meeting clinical criteria for Dravet syndrome. Crucially, 3 patients initially classified as having “unclassified seizures” were later re‐evaluated and recognized as having Dravet‐like features only after genetic diagnosis highlighting how molecular data can refine phenotypic classification. KCNQ2‐related neonatal epilepsy (n = 3) presented with FBTC, GTC or GC seizures in the first week of life. Two patients with WWOX p.Arg264* (Cases 29, 51) exhibited generalized epileptic spasms, developmental regression, and abnormal MRI—classic for WWOX‐associated DEE28. A particularly noteworthy finding was the identification of a novel homozygous missense variant in DYNC1H1 (Case 13, p.Lys1115Asn) in a neonate presenting with GT seizures and lissencephaly. While DYNC1H1 is traditionally associated with autosomal dominant neurodevelopmental disorders, our identification of a homozygous variant suggests a recessive inheritance pattern, thereby expanding the known allelic and clinical spectrum of the gene beyond its typical dominant manifestations. Two siblings with TBC1D23 frameshift variants (Cases 3–4) displayed corpus callosum agenesis, pontocerebellar hypoplasia, and focal impaired awareness seizures—consistent with Pontocerebellar Hypoplasia Type 11 (PCH11) but with earlier seizure onset than previously reported. Conversely, non‐DEE‐associated variants (e.g., PRRT2, MECP2, KMT2D) were often linked to milder or syndromic presentations, such as Rett‐like features, intellectual disability without regression, or self‐limited familial epilepsies. This phenotypic stratification supports the dual utility of NGS: not only for diagnosing monogenic DEEs but also for subtyping complex neurodevelopmental disorders where epilepsy is a secondary feature.

4.3. Clinical predictors of diagnostic yield

While intellectual disability (ID/DD) and drug resistance were highly prevalent (66% and 28%, respectively), neither significantly predicted a genetic diagnosis in our cohort (p > 0.05). This challenges the assumption that severe phenotypes uniformly increase diagnostic yield and suggests that phenotypic severity alone is a triage criterion for NGS.

The timing of seizure onset serves as a critical indicator for genetic etiology; neonatal and infantile onset are strongly associated with a higher likelihood of molecular diagnosis. A prospective national cohort study found that presentation before 6 months of age is significantly associated with a genetic diagnosis, with single‐gene disorders accounting for a quarter of seizure disorders in children presenting before 36 months. Meta‐analytical data further supports this, showing a pooled diagnostic yield of 29.3% specifically for the neonatal/infantile subgroup. ²⁸ Furthermore, clinical comorbidities such as intellectual disability (ID) and developmental delay (DD) significantly increase diagnostic success rates. While the general diagnostic yield in patients with autistic features is approximately 17.1%, this figure rises to 24.6% when autistic features co‐occur with ID or DD. ²⁹ These findings align with emerging evidence linking specific genes, such as PCDH19, to epilepsy–autism comorbidity, particularly in early childhood. Collectively, these data reinforce the concept that early‐onset epilepsies with neurodevelopmental features are frequently monogenic, warranting primary genetic investigation in these cohorts. ²⁹

Abnormal EEG findings were present in 59.4% (117/197) of patients with available data, and 45.3% (53/117) of these received a definitive diagnosis suggesting that epileptiform activity may serve as a phenotypic red flag for underlying genetic etiology. Of the 193 patients for whom imaging data were available, 99 (51.3%) had abnormal brain MRI. Structural abnormalities (e.g., white matter atrophy, matter changes, cortical dysplasia) were present in 43.8% (32/73) of patients with genetically diagnosed MRIs. This finding was not statistically significant in our analysis when compared to undiagnosed cases. Future studies with standardized neuroimaging protocols may better define this relationship.

Molecular diagnosis rates based on the clinical presentation of our cohort are presented graphically in Figure 3.

FIGURE 3.

Phenotypic characteristics and diagnostic rates of the entire cohort. ID/DD, intellectual disability/ developmental delay; ADHD, attention‐deficit/hyperactivity disorder (DSM‐5 criteria).; DEE, developmental and epileptic encephalopathy; *MRI data available for 193/250 (77.2%) and EEG data available for 197/250 (78.8%) patients.

4.4. The VUS conundrum and path to reclassification

A major challenge in clinical genomics remains the interpretation of variants of uncertain significance (VUS), identified in 22% of our cohort. While excluded from primary diagnostic yield calculations, many VUS occurred in genes with strong disease validity (CNTNAP2, GAD1, RORB, GRIN2A) and matched the patient's phenotype. For example, homozygous GAD1 splice variants were identified in two siblings with infantile‐onset tonic seizures finding biologically plausible given GAD1's role in GABA synthesis.

To address this, we propose a pragmatic, phenotype‐driven triage system to prioritize VUS for future re‐evaluation, specifically intended for research prioritization and not for immediate clinical diagnosis:

• Tier 1: VUS in genes with definitive ClinGen validity and strong phenotypic match → prioritize for targeted clinical surveillance, longitudinal follow‐up, and further evidence accrual (e.g., functional or segregation data) to clarify pathogenicity.

• Tier 2: Novel missense/splice variants in constrained genes (pLI ≥0.9) → prioritize for trio sequencing and segregation analysis. Genes with pLI ≥0.9 are highly intolerant to loss‐of‐function variants in population databases (e.g., gnomAD), indicating strong evolutionary constraint and increased likelihood that missense/splice variants in these genes are pathogenic. Trio sequencing is preferred over targeted parental segregation analysis for comprehensive variant detection—including de novo variants, compound heterozygosity, and uniparental disomy—but targeted segregation may be considered when resources are limited or a single high‐priority candidate variant exists.

• Tier 3: All other VUS entering longitudinal registries to enable future reclassification via data sharing (e.g., ClinVar, DECIPHER).

All VUS should undergo systematic re‐evaluation: Tier 1 variants at 6–12 months; Tier 2–3 variants at 2 years; and unresolved variants at ≥3 years or upon emergence of new functional evidence, additional affected family members, or significant phenotypic evolution.

Importantly, no VUS was used to alter clinical management in this cohort. Without trio analysis, de novo variant confirmation and segregation studies were limited, an acknowledged constraint that likely underestimates true diagnostic yield.

4.5. Precision medicine impact

As detailed in Section 3.5, genetic diagnosis directly informed therapeutic decisions in 12 of 89 diagnosed patients (13.5%): SCN1A‐positive cases (n = 9) avoided sodium channel blockers to reduce seizure exacerbation risk; SLC2A1 variants (n = 2) prompted ketogenic diet trials with documented seizure reduction; and MECP2 diagnosis (n = 1) enabled Rett syndrome‐specific surveillance. These outcomes mirror recent studies showing NGS‐driven management changes in 30–50% of pediatric epilepsy cases. ³⁰ , ³¹ From a health economics perspective, early genomic diagnosis may reduce unnecessary diagnostic tests, hospitalizations, and ineffective therapies, though cost‐effectiveness analyses in Turkish settings remain needed.

4.6. Limitations and future directions

This study has key limitations:

1. No CNV or noncoding variant analysis: the absence of CNV analysis potentially limits diagnostic success, as combined detection of sequence and intragenic copy number variants has been shown to provide the sole molecular diagnosis in 5.1% of patients, significantly increasing the overall yield. ³²

2. The unavailability of biological parental samples for trio analysis necessitated a proband‐only interpretation, thereby limiting de novo variant confirmation and segregation studies.

3. EEG and MRI data were not uniformly available, reducing power for neurophysiological correlations.

4. Functional validation (e.g., electrophysiology for SCN1A variants) was not performed.

Future prospective studies integrating trio‐based genome sequencing, RNA‐seq for splice variant validation, and systematic deep phenotyping using Human Phenotype Ontology (HPO) terms are warranted to resolve VUS, uncover noncoding variants, and ultimately increase diagnostic yield beyond the 57.6% threshold observed here.

AUTHOR CONTRIBUTIONS

Kadri Karaer: Conceptualization; methodology; supervision; data curation; writing—original draft; writing—review and editing. Derya Karaer: Data curation; formal analysis; investigation; writing—review and editing. Beste Kipçak Yüzbaşı: Investigation; resources; clinical phenotyping. İpek Şahin: Investigation; data curation; clinical phenotyping. Olcay Güngör: Investigation; clinical validation.

FUNDING INFORMATION

This research did not receive any specific grant from funding agencies in the public, commercial, or not‐for‐profit sectors.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no competing interests.

PATIENT CONSENT

Written informed consent was obtained from the legal guardians of all participants for genetic testing and the use of anonymized clinical and genomic data for research purposes.

Test yourself.

1. According to the findings in this pediatric epilepsy cohort, which of the following clinical features is a statistically significant predictor for establishing a definitive genetic diagnosis using next‐generation sequencing (NGS)?

   1. Drug‐resistant epilepsy
   2. Presence of intellectual disability/developmental delay (ID/DD)
   3. Neonatal seizure onset
   4. Abnormal structural findings on brain MRI
2. The study identified the PRRT2 c.649dup frameshift variant in four unrelated families and noted a high proportion (26%) of autosomal recessive diagnoses. Which underlying demographic factor in the Turkish population most strongly contributes to this distinct genetic architecture?

   1. High prevalence of de novo mutations
   2. High rates of consanguineous marriages
   3. A predominant X‐linked inheritance pattern
   4. A low diagnostic efficiency of Whole Exome Sequencing (WES)
3. Identifying a definitive molecular etiology can directly alter clinical management (clinical utility). In this study, what was the primary, actionable therapeutic change for patients diagnosed with SCN1A pathogenic variants?

   1. Initiation of ketogenic diet therapy
   2. Avoidance or discontinuation of sodium channel blockers
   3. Evaluation for epilepsy surgery
   4. Prescription of potassium channel openers

Answers may be found in the Supporting Information

Supporting information

Appendix S1.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request. The raw genomic data are not publicly available due to privacy and ethical restrictions involving human participants.