Combined family-based association and linkage analyses in families affected by attention-deficit hyperactivity disorder Authors: Cristina M. Justice, Kwangmi Ahn, Benjamin Jung, Luke J. Norman, Gustavo Sudre, Wendy Sharp, Marine Bouyssi-Kobar, Saadia Choudhury, Stevi Gligorovic, Paul Kundzicz, Maria T. Acosta, Philip Shaw Published: 2026 ABSTRACT Attention deficit hyperactivity disorder (ADHD) is one of the most prevalent and heritable of neurodevelopmental disorders. To characterize the genetic variants contributing to this heritability, we studied families with members affected by ADHD. Genome-wide array data were obtained on two cohorts: NHGRI Family Cohort (359 nuclear families,1538 individuals) and NCR Family Cohort (25 multigenerational and 132 nuclear families, 631 members). A meta-analysis of family-based association tests (FBAT) of the two cohorts identified three genome-wide significant associations, one upstream from TFRC, and two that were intronic to GSDME and to TRIM31, with the last gene previously implicated in ADHD. Genes associated with ADHD (MAGMA, gene-based association P < 0.05) overlapped with genes implicated in ADHD by prior case/control genome-wide association studies. Meta-analyses of the linkage signals identified one significant linkage region (LOD > 3) on 19p13.2-13.11 that encompassed a genome-wide significant variant in the FBAT of the NHGRI Family Cohort (rs55741253, P = 2.68 × 10− 8). Additionally, two of the five linkage regions that reached a LOD > 2 in our meta-analysis replicated significant linkages from prior studies (prior studies reporting 13 linkages with LOD > 3.0; hypergeometric test of overlapping linkage signals, P < 0.01). Using neuroimaging data from the NCR cohort, we found genes associated with ADHD overlapped with genes associated with the brain’s total surface area and a feature of functional connectivity, reflecting the interaction between the brain’s default mode and task positive networks. Such work begins to delineate the neural substrates of the discovered genetic associations with familial ADHD.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00439-026-02840-7. Introduction Attention-deficit/hyperactivity disorder (ADHD) is a prevalent psychiatric disorder characterized by impairing symptoms of inattention, hyperactivity and impulsivity with heritability estimates of between 77 and 88% (Faraone and Larsson 2019). Recent landmark discoveries into common genetic variants conferring risk for ADHD have been made through genome-wide association studies (GWAS) from the Psychiatric Genomics Consortium of unrelated ADHD cases and unaffected controls (Demontis et al. 2019; 2023). Here we extend gene discovery through applying family-based association and linkage analyses to individuals from nuclear or extended multigenerational families affected by ADHD. Our family-based approach is appropriate given the high heritability of ADHD and is more resistant to issues of population stratification than those present in case-control studies of unrelated individuals (Lange et al. 2008). By identifying genomic variation associated with familial ADHD, the study can thus both complement and extend the recent GWAS findings on ADHD, in which cases were ascertained without reference to the presence or absence of family history. Prior studies of familial ADHD have mainly used linkage analyses, mapping the segregation of ADHD with genetic loci across generations. Five studies have reported significant linkage at LOD > 3.0 for three genomic locations (5p13 (Ogdie et al. 2004; Hebebrand et al. 2006), 9q22 (Asherson et al. 2008; Romanos et al. 2008) and 17p11 (Ogdie et al. 2004; Arcos-Burgos et al. 2004) with independent replication for all signals occurring only at nominal levels of significance (LOD > 1.5). A meta-analysis data combining the results from four projects, for a total of 2,064 trios and 896 cases and 2,455 controls, did not find any genome-wide significant associations.(Neale et al. 2010). Several ADHD studies have used a family-based association test approach. The largest involved 958 trios with probands diagnosed with ADHD and conduct disorder and reported no variants reaching genome-wide significance (Anney et al. 2008), while another study of 607 families found nominal significant to 12 SNPs (Oades et al. 2008). Here we use both approaches- family-based association tests and linkage analyses - to examine genetic associations in two family cohorts. The first cohort, the NHGRI Family Cohort, comprises 1,538 individuals from 359 nuclear families (Acosta et al. 2008), including 423 individuals with ADHD and 1,115 unaffected individuals. The second cohort, the NCR Family Cohort, comprises 631 individuals from 25 multigenerational extended families and 132 nuclear families, including 239 individuals with ADHD and 392 unaffected individuals. Both cohorts were ascertained through families with ADHD and therefore are not representative of the general population. For context, the global prevalence of ADHD has been estimated at 7.6% in children and 5.6% in adolescents based on cross-sectional studies worldwide, which is substantially lower than the proportion observed in our cohorts due to this ascertainment strategy. Linkage and family-based association approaches can provide complementary insights: for example, while linkage analyses detect peaks that can encompass many genes, family-based associations falling within a peak can help prioritize loci or genes for further characterization. Equally, findings from linkage and family-based associations can also differ as they calculate the relationship between phenotype and genotype differently. Once genetic associations with familial ADHD are ascertained it becomes important to consider how they might be acting on brain structure and function to give rise to symptoms. One of the cohorts in the current study had multimodal neuroimaging imaging available on many individuals, affording an opportunity to begin this process. Here, we focused on neural differences that have been robustly associated with ADHD, through mega-analytic, large scale imaging studies (Hoogman et al. 2017; 2019; Norman et al. 2022; Sudre et al. 2022). Leveraging in vivo neuroimaging, we considered both the anatomic changes most consistently tied to ADHD, and anomalies in the brain’s structural and intrinsic functional connectivity, based on a model of ADHD as being tied to disrupted interactions between key brain functional networks (Norman et al. 2022; Sudre et al. 2022). We estimated the overlap between genes associated with familial ADHD and genes associated with these key neural features. Such work is a first step towards parsing the neural substrates that may help explain associations between common genetic variation and the ADHD symptoms. In summary, we report on family-based association and linkage analyses on two family cohorts, comprising a total of 2,169 individuals. We determine if genes emerging in our family-based study overlapped with those implicated through recent GWAS of unrelated cases and controls. Finally, we leverage neuroimaging data to begin to parse the neural substrates that might contribute to the discovered genetic associations. Although this is a small sample, compared to the number of individuals used in recent case/control studies of ADHD, it consists of linkage and association tests on a large number of extended families, which are known to find variants with larger effects than those identified through case/control associations. Materials and methods Cohorts NHGRI family cohort The main inclusion criterion for families was the presence of a child with ADHD aged between 7 and 17 years old, with at least one sibling, either affected or unaffected, ideally with both biological parents available for assessment (Acosta et al. 2008). Families with two affected parents were excluded, because such pedigrees complicate interpretation of transmission patterns and reduce clarity in distinguishing parental contribution within family-based analyses. The comorbidities listed in Table S1 were excluded since they can confound the diagnosis and introduce alternative biological causes of inattention and hyperactivity. Diagnosis of childhood ADHD was based on the clinician interview with the parent (Diagnostic Interview for Children and Adolescents-IV [DICA]) (Reich 2000). Assessment of ADHD in adults (parents and siblings) aged over 18 years was based on the Vanderbilt ADHD Diagnostic Rating Scale (VADRS) (Wolraich et al. 2003), using the cut-off of > 5 on symptom counts. For adults, a retrospective diagnosis of probable childhood ADHD was defined based on scores from the WURS (score > 35 if female and > 40 if male). Only one comorbidity, childhood oppositional defiant disorder (ODD), was sufficiently prevalent to be considered in separate FBAT analyses. Principal component analysis (PCA) performed using PLINK v1.9.0 (Purcell et al. 2007) identified 90.6% of the individuals in this cohort as white non-Hispanic (WNH). PCA analysis was performed to characterize this cohort. No individuals were excluded based on the PCA results, as the family-based tests used in this study were robust to population stratification. The study was approved by the institutional review board of the NHGRI; written informed consent was obtained from adult participants and parents. DNA was extracted from blood and genotyped in two phases. First, 840 DNA samples consisting mainly of parent offspring trios (481 males and 359 females) were genotyped with the Global Screening Array v1.0 with Multi-Disease Component (GSA-MD v1.0) at the Broad Stanley Center for Psychiatric Disease, Broad Institute, Cambridge, MA, USA. An additional 775 samples (439 males and 336 females), consisting primarily of discordant and concordant siblings, as well as 8 replicates and 81 samples previously genotyped with the GSA v1.0, were genotyped at the Genomic Core of the NHGRI, using the GSA v3.0. The samples genotyped on GSAv1.0 and GSAv3.0 were merged, resulting in 601,123 overlapping variants available for analysis. Map position information was based on GRCh37/hg19. Any variants that were discordant among 81 replicates genotyped on both arrays were discarded. Variants were also dropped if missing rate > 5%, were unmapped, had a Mendelian inconsistency in > 1 trio, had minor allele frequency (MAF) < 5%, deviated from Hardy-Weinberg equilibrium (HWE) (P < 0.00001) or were discordant between duplicates. The genotypes for variants with a Mendelian inconsistency found in only one trio were zeroed out. Identity-by-descent (IBD) and Mendelian error analyses were conducted with PLINK v1.9.0 (Purcell et al. 2007). For the IBD analysis, we used a panel of variants that were not in linkage disequilibrium, and which were retained after all quality control thresholds previously mentioned were met. Four monozygotic (MZ) twins and one set of identical triplets were identified, of which one individual from each set with the highest genotyping rate was retained. MZ twins are almost identical at every variant, whether these are disease-associated or not, so keeping both MZ twins in a linkage study can lead to false positives. A total of 55 samples were removed due to IBD and/or Mendelian errors. The remaining merged set of individuals genotyped with GSAv1.0 and GSAv3.0 consisted of 1,538 individuals (877 males and 661 females) belonging to 359 informative families (Table 1). Of these 359 families, 330 met the proposal goal of enrolling a child with ADHD and at least one affected or unaffected sibling, while 29 families were composed of case-parent trios, for which the additional sibling did not pass quality control filtering. Table 1Gender, age, and comorbidity information of the NHGRI Family and NCR Family Cohorts analyzed using FBATNHGRI FamilyNCRTotal1,538631Males (%)877 (57.0)363 (57.5)Females (%)661 (43.0)268 (42.5)Average age (standard deviation)28 (17)25 (18)Number of families359157Median family size (range)4.28 (3–18)4.02 (2–25)ADHD (%)423 (27.5)239 (37.9)Race (%)White1442 (93.8)513 (81.3)African American23 (1.5)49 (7.8)Asian26 (1.7)21(3.3)Native American10 (0.7)> 1 Race37 (2.4)38 (6.0)Other (%)3 (0.5)Not specified (%)7 (1.1)Ethnicity (%)Hispanic58 (3.8)49 (7.8)Non-Hispanic1480 (96.2)579 (91.8)Not specified3 (0.5)Comorbidities (%) for individuals < 19 years of ageOppositional Defiant Disorder (ODD)205 (28.7)25 (7.0)Conduct Disorder16 (2.2)1 (0.3)Major Depressive Disorder4 (0.6)16 (4.3)Mania2 (0.3)Dysthymic Disorder8 (1.1)Separation Anxiety5 (0.7)1 (0.3)Generalized Anxiety25 (3.5)12 (3.3)Obsessive Compulsive Disorder11 (1.5)1 (0.3)Post-traumatic stress disorder2 (0.3)ADHD + ODD diagnosis184 (42.8)31 (8.5)Comorbidities (%) for individuals ≥ 19 years of ageScreened Positive for Anxiety or Depression116 (14.5)53 (19.0)Dysthymic Disorder1 (0.5) Table 1Gender, age, and comorbidity information of the NHGRI Family and NCR Family Cohorts analyzed using FBATNHGRI FamilyNCRTotal1,538631Males (%)877 (57.0)363 (57.5)Females (%)661 (43.0)268 (42.5)Average age (standard deviation)28 (17)25 (18)Number of families359157Median family size (range)4.28 (3–18)4.02 (2–25)ADHD (%)423 (27.5)239 (37.9)Race (%)White1442 (93.8)513 (81.3)African American23 (1.5)49 (7.8)Asian26 (1.7)21(3.3)Native American10 (0.7)> 1 Race37 (2.4)38 (6.0)Other (%)3 (0.5)Not specified (%)7 (1.1)Ethnicity (%)Hispanic58 (3.8)49 (7.8)Non-Hispanic1480 (96.2)579 (91.8)Not specified3 (0.5)Comorbidities (%) for individuals < 19 years of ageOppositional Defiant Disorder (ODD)205 (28.7)25 (7.0)Conduct Disorder16 (2.2)1 (0.3)Major Depressive Disorder4 (0.6)16 (4.3)Mania2 (0.3)Dysthymic Disorder8 (1.1)Separation Anxiety5 (0.7)1 (0.3)Generalized Anxiety25 (3.5)12 (3.3)Obsessive Compulsive Disorder11 (1.5)1 (0.3)Post-traumatic stress disorder2 (0.3)ADHD + ODD diagnosis184 (42.8)31 (8.5)Comorbidities (%) for individuals ≥ 19 years of ageScreened Positive for Anxiety or Depression116 (14.5)53 (19.0)Dysthymic Disorder1 (0.5) Gender, age, and comorbidity information of the NHGRI Family and NCR Family Cohorts analyzed using FBAT NCR family cohort The Neurobehavioral Clinical Research (NCR) Family Cohort comprised both extended and nuclear families- demographic details in Table 1. The inclusion criteria for the extended families were (1) the presence of second-, third-, or higher-degree relatives and (2) a diagnosis of ADHD in at least 25% of family members. For nuclear families, the main inclusion criterion was at least two siblings, one with ADHD. The study was approved by the institutional review board of the NHGRI; written informed consent was obtained from adult participants and parents; children gave written assent. For children, diagnosis of ADHD was through the parental Diagnostic Interview for Children and Adolescents-IV (DICA) (Reich 2000). For adults, the Conners’ Adult ADHD Diagnostic Interview for DSM-IV(CAADID) (Epstein, Johnson, and Conners 2001; Conners and Staff 2004) was used to ascertain both adult and childhood history of ADHD symptoms. Presence of other psychiatric diagnoses was established through the Structured Clinical Interview for DSM-IV-TR Axis I Disorders, Research Version, Patient Edition (First et al. 2002). All interviews were conducted by 2 experienced clinicians (W.S. and P.S.) with interrater reliabilities of κ > 0.9. The main exclusion criteria were an IQ less than 70 (Wechsler 2011), neurologic disorders affecting brain structure, or psychotic disorders. Further details on data collection, genotyping process and phenotype information collected are detailed elsewhere (Sudre et al. 2017, 2020, 2021). The samples were genotyped on the Infinium OmniExpressExome-8v1. Quality control steps included dropping variants if genotyping call rate < 95%, were not mapped, had MAF < 5%, were discordant in duplicates (21 duplicates) or HWE P < 0.00001. Mendelian inconsistencies were detected using PLINK v1.9.0 (Purcell et al. 2007) and PEDSTATS (Wigginton and Abecasis 2005). Variants with > 1 Mendelian inconsistency were dropped; if inconsistency found in only one trio, the genotypes for corresponding variant and trio were zeroed out. IBD analyses were conducted with PLINK v1.9.0 (Purcell et al. 2007). Eighteen monozygotic twins were identified, of which the twin from each pair with the higher missing genotyping rate was dropped. One family and one sib-pair were dropped due to IBD inconsistencies. For the duplicated samples, we dropped one with higher missing genotyping rate. The final sample consisted of 631 individuals (268 females and 363 males) from 157 families (Table 1), of which 73.2% were identified as WNH though PCA using PLINK v1.9.0 (Purcell et al. 2007). No samples were excluded based on PCA results, since the family-based tests used are immune to population stratification bias. For the NCR Family Cohort only, neuroanatomic data (acquired using T1-weighted images on a 3 T Siemens scanner) was available on 370 subjects that passed quality control. We extracted neuroanatomic features that have been most consistently associated with ADHD through multi-site mega-analytic, cross-sectional studies: total brain volume, striatal and amygdala volume reduction (Hoogman et al. 2017), right lateral orbitofrontal and superior frontal cortical surface area reductions, and a thinner right fusiform and precentral cortex (Hoogman et al. 2019). In addition, diffusion tensor imaging that passed quality control was available on 345 subjects from which we extracted measures of the microstructure (fractional anisotropy) of the two tracts (inferior longitudinal fasciculi and left uncinate fasciculus) that were robustly associated with ADHD in a recent mega-analysis (Sudre et al. 2022). Finally, 359 subjects had functional connectivity metrics from resting state fMRI- from which we extracted a measure of the atypical connectivity between the default mode network (that characterizes interoceptive processing) and networks supporting attention, cognitive control and motor planning and execution, as we found these interactions to be most robustly associated with ADHD (Norman et al. 2022). Details of acquisition and quality control in Supplemental Methods. There was thus a total of 20 neural measures. Imputation analysis The two cohorts were genotyped on different arrays with less than 100,000 shared variants. To increase overlap the cleaned genotypes were fed to the Mimimac4 software (Das et al. 2016) on the University of Michigan Imputation server v.1.2.4 (https://imputationserver.sph.umich.edu), using the 1000G Phase 3 v5 GRCh37/hg19 reference panel and the European (EUR) population. Genotypes were phased prior to imputation using Eagle v2.4 (Das et al. 2016) and filtered by imputation quality score Rsq > 0.3. Imputed variants were dropped prior to analyses if they had an HWE P < 0.00001, MAF < 5%, or any Mendelian inconsistency detected by using either PLINK v1.9.0 (Purcell et al. 2007) or PEDSTATS (Wigginton and Abecasis 2005)). For association and linkage analyses, we selected for analysis the 244,195 variants initially genotyped in NHGRI Family Cohort that had either genotyped or imputed corresponding variants in the NCR Family Cohort that passed quality control. Family-based association test (FBAT) The GWAS was carried out using the Family-Based Association Test (Rabinowitz and Laird 2000; Laird et al. Laird 2000). The FBAT is a generalization of the TDT (Spielman, McGinnis, and Ewens 1993), which measures the rate of transmission of an allele from the heterozygous parents to an affected offspring. FBAT (v2.0.2) breaks the extended pedigrees into nuclear families and can deal with missing parental genotypes and phenotypes, while testing for associations in the presence of linkage. Association results are the sum of the statistics across the independent families, assuming an additive model with a null hypothesis of no linkage or association. The minimum number of informative families used for calculating the test statistic was set at 10. FBAT was run independently on the 359 families in the NHGRI Family Cohort and the 157 families from the NCR Family Cohort. We performed a meta-analysis of summary statistics from FBAT of the two cohorts with METAL (http://www.sph.umich.edu/csg/abecasis/metal/index.html) using a fixed-effects meta-analysis weighted by sample number for each study and testing for between-study heterogeneity. A heterogeneity p-value (Phet) > 0.1 indicated homogeneity of the effect size across both cohorts. Linkage analysis Multipoint non-parametric linkage analysis was performed using MERLIN v1.1.2 (Abecasis et al. 2002) to test for linkage between the diagnosis of ADHD and the array variants. The genetic distances (in cM) were assumed to be a linear function of the map distance (in Mb), and the physical distances were converted to map units (1 cM = 1 Mb). Prior to linkage analysis, variants were pruned for linkage disequilibrium (LD) with PLINK v1.9.0 (Purcell et al. 2007) using a pairwise analysis of 5 variants at a time in a 50 kb window and a pairwise r2 threshold < 0.2. This resulted in 62,239 autosomal variants with an MAF > 5% available for analysis. Only families with two parents and at least two offspring (380 families, 1831 individuals) were analyzed for linkage, of which 330 families (1451 individuals; family size range: 4–18) were from the NHGRI Family Cohort and 50 families (380 individuals; family size range: 4–18) were from the NCR Family Cohort. Allele frequencies were estimated from the founders. The non-parametric linkage statistic p-values for Whittemore and Halpern (Whittemore and Halpern 1994), and the Kong and Cox logarithm of odds (LOD) scores (Kong and Cox 1997) were calculated for each variant, as implemented in MERLIN (Abecasis et al. 2002). The NHGRI Families and NCR cohorts were analyzed separately. A meta-analysis of the cohorts, using the Whittemore and Halpern p-values, was performed using METAL(http://www.sph.umich.edu/csg/abecasis/metal/index.html). A LOD score > 3 was used as the threshold for significant evidence of linkage, while a LOD > 2 was selected as suggestive evidence of linkage, with the candidate region set at ± 1 mega base (Mb) from the linkage peaks. Although more stringent thresholds have been proposed for genome-wide linkage analyses (e.g., LOD ≥ 3.3; Lander and Kruglyak 1995), these thresholds may be too conservative for complex traits and studies with smaller sample sizes. Sawcer et al. (1997), using simulations based on their specific genome-wide linkage scan, estimated that a LOD score of 3.2 corresponded to a genome-wide significance level of 5%, demonstrating that significance thresholds depend on study design. Here we applied a LOD threshold of > 3 to balance statistical power with control of false positives. There was a total of 13 candidate linkage regions with LOD > 3.0 reported in the literature, see Table S3. The significance of the overlap of linkage signals from our meta-analysis with the previously reported linkage regions was determined by calculating the hypergeometric probability (https://molbiotools.com/math_calculators/hypergeometric.html), setting the total population size to 400, equal to the number of chromosomal bands in the genome (Rosenberg and Rosenberg 2012). SNPs annotation We performed ANNOVAR (http://annovar.openbioinformatics.org/) (Wang et al. 2010) gene-based annotation using Ensembl genes to annotate SNPs within FUMA (Watanabe et al. 2017). Gene-based association and gene set enrichment analysis We conducted gene-based association analyses using the summary statistics from the FBAT of the larger NHGRI Family Cohort and from the meta-analysis of both cohorts. MAGMA v.1.08 (de Leeuw et al. 2015) (http://ctg.cncr.nl/software/magma) was used to assess the aggregated genetics effects within a locus across all protein-coding genes in the genome, declaring exome-wide significant genes at P < 3.08 × 10− 6 (0.05/16192 genes). Gene length boundaries were defined as 10 kilobases (kb) upstream and downstream from start and stop site, respectively, to include regulatory elements. The NCBI 37.3 build was used to assign the genetic variants to each gene. Genes associated with ADHD, either through FBAT or by MAGMA gene-based association to conduct gene-set enrichment analyses, were fed into MAGMA’s competitive gene-set analysis facility using the GENE2FUNC module integrated in FUMA. Gene sets in MAGMA were obtained from the MSigDB v.7.0 for “Curated gene sets” and “GO terms”, and from reported genes from the GWAS-catalog. Gene set enrichment analysis were also conducted using the R software package WebGestaltR (0.4.3) for Biological Process, Molecular Function, and Cellular Components. In addition, we created curated gene-sets from GWAS and transcriptome-wide association study (TWAS) for ADHD, autistic spectrum disorder, schizophrenia, and bipolar affective disorder (curated gene sets listed in Table S2). The enrichment analysis was performed using the hypergeometric test. Input was the summary statistics from MAGMA gene-based associations. For both sets of gene-set enrichment analyses, false discovery rate (FDR)-adjusted q values were generated using Benjamini and Hochberg’s method to account for multiple testing. Using the GeneOverlap R package (v3.12), Fisher’s exact test was tested for overlap between genes associated with the clinical and with the brain phenotypes in the NCR Family Cohort. Polygenic risk score (PRS) analyses We assessed if polygenic risk scores (PRS) for ADHD, derived from GWAS of unrelated cases and controls (the summary data from the Psychiatric Genetics Consortium (PGC) ADHD GWAS (Demontis et al. 2019)) was also associated with ADHD within our family cohorts. For this analysis, we focused on only samples of European ancestry based on genetic principal components (PC-Air (Conomos, Miller, and Thornton 2015)). We used SNP effect sizes estimates (20,183 ADHD cases and 35,191 control) from PGC summary data (https://pgc.unc.edu). Given the genetic overlap between ADHD and other disorders, we also considered polygenic risk for autistic spectrum disorders, obsessive compulsive disorder, Tourette’s syndrome, bipolar affective disorder and schizophrenia, derived from summary GWAS data from the PGC (Grove et al. 2019; International Obsessive Compulsive Disorder Foundation Genetics and Studies 2018; Yu et al. 2019; Mullins et al. 2021; Pardinas et al. 2018). After clumping, PRS were generated using PRSice-2 v2.3.5 (Choi and O’Reilly Choi 2019) for each sample at P value thresholds of: 1, 0.5, 0.05, 0.04, 0.03, 0.02. 0.01, 0.005, 0.001, 1 × 10− 4, 1 × 10− 5, 5 × 10− 8. All PRS values were normalized before analyzing. The association between PRS and ADHD status was then tested using a generalized linear mixed model implemented in the GMMAT package (Chen et al. 2016). In this model, PRS was included as a fixed-effect predictor along with age, sex, and ancestry principal components, while family structure was modeled as a random effect using a genetic relationship matrix (GRM) calculated from pruned genotype data using PLINK v1.9.0 (Purcell et al. 2007). The glmm.wald function was used to perform association testing. The number of principal components retained for each cohort was determined based on the proportion of genetic variance explained, with the minimum number of PCs selected to capture at least 85% of the variance (resulting in five PCs for the NHGRI Family Cohort and three for the NCR Family Cohort). To account for multiple testing across the p-value thresholds evaluated within each PRS analysis, false discovery rate (FDR)-adjusted p-values were calculated using the Benjamini–Hochberg procedure. Results Association results We considered 1,538 individuals [877 males; 661 females; mean age 28 (17 SD); age range 4–78] from 359 families [median size 9.5, range 3–18] in the NHGRI Family Cohort (Table 1). Nineteen SNPs were associated with ADHD at a genome-wide level of significance P < 5 × 10− 8 (Figure S1a, Table 2). Table 2Genome-wide significant (P < 5 × 10−8) results for association analysis of the NHGRI family cohortCHRBPSNPA1A2MAFP-valuesLocationGene3195,838,160rs77176301CT0.0592.95E-11upstreamLINC00885, TFRC4111,707,130rs62337211AC0.156.17E-10intergenicPITX2, MIR2974154,603,296rs2405433TC0.191.87E-13upstream; downstreamTLR2, LOC100419170, RNF175, TMEM131L630,036,333rs17187693CG0.294.81E-08intronic PPP1R11 630,074,086rs115147572AC0.111.98E-10intronic TRIM31 631,045,554rs9378152GC0.134.43E-08upstream; downstreamPSORS1C1, C6orf15, CDSN, HCG22, MUC2277,223,618rs60600046GA0.061.52E-10Intronic C1GALT1 724,783,987rs61241329AC0.075.95E-10intronic GSDME 877,927,157rs536249779TG0.127.81E-10upstreamMIR3149, PEX2914,246,998rs10961448CA0.084.25E-08intronic NFIB 102,742,533rs7911061AG0.162.22E-08intergenicLINC02645, LOC1019278241452,778,656rs5003100AG0.081.92E-13upstream; downstreamPTGER2, PTGDR1493,263,982rs17128572CG0.164.58E-10exonic GOLGA5 1859,994,265rs8087597GA0.153.82E-10intronic TNFRSF11A 1860,024,128rs58112300CT0.061.79E-09intronic TNFRSF11A 1939,993,470rs55741253TA0.242.68E-08exonic DLL3 206,697,162rs34169461TA0.114.80E-13downstream LINC01713 208,817,556rs58568270GA0.15.61E-15intronic PLCB1 2135,715,023rs73902815CT0.061.36E-11upstreamKCNE2, SMIM11A, SMIM11BCHR chromosome, BP base pair position (GRCh37/hg19), A1 minor allele in NHGRI Family Cohort, A2 major allele in NHGRI Family Cohort, MAF1 minor allele frequency estimated from 1106 NHGRI Family Cohort founders Table 2Genome-wide significant (P < 5 × 10−8) results for association analysis of the NHGRI family cohortCHRBPSNPA1A2MAFP-valuesLocationGene3195,838,160rs77176301CT0.0592.95E-11upstreamLINC00885, TFRC4111,707,130rs62337211AC0.156.17E-10intergenicPITX2, MIR2974154,603,296rs2405433TC0.191.87E-13upstream; downstreamTLR2, LOC100419170, RNF175, TMEM131L630,036,333rs17187693CG0.294.81E-08intronic PPP1R11 630,074,086rs115147572AC0.111.98E-10intronic TRIM31 631,045,554rs9378152GC0.134.43E-08upstream; downstreamPSORS1C1, C6orf15, CDSN, HCG22, MUC2277,223,618rs60600046GA0.061.52E-10Intronic C1GALT1 724,783,987rs61241329AC0.075.95E-10intronic GSDME 877,927,157rs536249779TG0.127.81E-10upstreamMIR3149, PEX2914,246,998rs10961448CA0.084.25E-08intronic NFIB 102,742,533rs7911061AG0.162.22E-08intergenicLINC02645, LOC1019278241452,778,656rs5003100AG0.081.92E-13upstream; downstreamPTGER2, PTGDR1493,263,982rs17128572CG0.164.58E-10exonic GOLGA5 1859,994,265rs8087597GA0.153.82E-10intronic TNFRSF11A 1860,024,128rs58112300CT0.061.79E-09intronic TNFRSF11A 1939,993,470rs55741253TA0.242.68E-08exonic DLL3 206,697,162rs34169461TA0.114.80E-13downstream LINC01713 208,817,556rs58568270GA0.15.61E-15intronic PLCB1 2135,715,023rs73902815CT0.061.36E-11upstreamKCNE2, SMIM11A, SMIM11BCHR chromosome, BP base pair position (GRCh37/hg19), A1 minor allele in NHGRI Family Cohort, A2 major allele in NHGRI Family Cohort, MAF1 minor allele frequency estimated from 1106 NHGRI Family Cohort founders Genome-wide significant (P < 5 × 10−8) results for association analysis of the NHGRI family cohort CHR chromosome, BP base pair position (GRCh37/hg19), A1 minor allele in NHGRI Family Cohort, A2 major allele in NHGRI Family Cohort, MAF1 minor allele frequency estimated from 1106 NHGRI Family Cohort founders The NCR Family Cohort consisted of 631 individuals [363 males; 268 females; mean age 25 (18 SD); age range 5–88] in 157 families [25 multigenerational, extended and 132 nuclear families, median family size 4, range 2–25 members]. No genome-wide significant associations with ADHD emerged in the NCR Family Cohort and no genome-wide significant variant in the NHGRI Family Cohort was significantly associated with ADHD in the NCR Family Cohort (Figure S1b). However, given the similarity in study design, we combined the cohorts meta-analytically, retaining the 258,399 overlapping variants that survived quality control procedures. Three of the 19 SNPs that had genome-wide significant associations with ADHD in the NHGRI Family Cohort also attained genome-wide significance (P < 5 × 10− 8) in the meta-analysis (Fig. 1): rs77176301, upstream from TFRC; rs115157572, intronic to TRIM31; and rs61241329, intronic to GSDME. In addition, twelve of the significant SNPs in the NHGRI Family Cohort were borderline significant in the meta-analysis (P < 1 × 10− 5) (Table 3, Figure S2), of which eight had homogeneous effects in both cohorts (Phet > 0.1). Several of the variants in Table 3 show evidence of heterogeneity (Pₕₑₜ < 0.05), indicating differences in effect sizes between cohorts. Fig. 1Manhattan and quantile-quantile (QQ) plots meta-analysis of FBAT results of NGRHI Family and NCR Family Cohorts. The horizontal blue line corresponds to P < 1 × 10− 5, while the horizontal red line corresponds the genome-wide significant threshold of P < 5 × 10− 8 Manhattan and quantile-quantile (QQ) plots meta-analysis of FBAT results of NGRHI Family and NCR Family Cohorts. The horizontal blue line corresponds to P < 1 × 10− 5, while the horizontal red line corresponds the genome-wide significant threshold of P < 5 × 10− 8 Table 3Meta-analysis results (P < 1 × 10−5) of the NHGRI family and NCR family cohortsCHRBPSNPA1A2MAF1MAF2 P P het FunctionGene3195,838,160rs77176301CT0.060.052.32E-100.021intergenicTFRC, LINC00885415,934,723rs2532098TG0.350.383.32E-060.866intergenicCD38, FGFBP1630,036,333rs17187693CG0.30.067.66E-070.0194intronic PPP1R11 630,074,086rs115147572AC0.110.071.47E-080.0037intronic TRIM31 644,160,786rs9462975CA0.080.064.83E-070.4503intergenicCAPN11, MYMX724,783,987rs61241329AC0.070.054.59E-080.0034intronic GSDME 877,478,539rs28727938CG0.170.061.72E-060.0593intergenicLINC01111, ZFHX4-AS110121,314,587rs116618532GA0.070.059.41E-060.3783intergenicRGS10, TIAL11174,137,551rs612040AC0.450.462.94E-060.1356intergenicMIR548AL, KCNE31174,153,008rs10793085TC0.460.478.59E-060.2894intergenicMIR548AL, KCNE311126,421,415rs59826922AG0.060.085.65E-060.7473intronic KIRREL3 1654,032,493rs11646260GA0.060.055.41E-060.0353intronic FTO 192,558,131rs78619038GA0.180.215.38E-060.1617intronic GNG7 1939,993,470rs55741253TA0.240.091.84E-060.0031exonic DLL3 2246,614,274rs1800206CG0.160.061.53E-060.4838exonic PPARA CHR chromosome, BP base pair position (GRCh37/hg19), A1 minor allele in NHGRI Family Cohort, A2 major allele in NHGRI Family Cohort, MAF1 minor allele frequency estimated from 1106 NHGRI Family Cohort founders, MAF2 minor allele frequency estimated from 208 NCR Cohort founders Table 3Meta-analysis results (P < 1 × 10−5) of the NHGRI family and NCR family cohortsCHRBPSNPA1A2MAF1MAF2 P P het FunctionGene3195,838,160rs77176301CT0.060.052.32E-100.021intergenicTFRC, LINC00885415,934,723rs2532098TG0.350.383.32E-060.866intergenicCD38, FGFBP1630,036,333rs17187693CG0.30.067.66E-070.0194intronic PPP1R11 630,074,086rs115147572AC0.110.071.47E-080.0037intronic TRIM31 644,160,786rs9462975CA0.080.064.83E-070.4503intergenicCAPN11, MYMX724,783,987rs61241329AC0.070.054.59E-080.0034intronic GSDME 877,478,539rs28727938CG0.170.061.72E-060.0593intergenicLINC01111, ZFHX4-AS110121,314,587rs116618532GA0.070.059.41E-060.3783intergenicRGS10, TIAL11174,137,551rs612040AC0.450.462.94E-060.1356intergenicMIR548AL, KCNE31174,153,008rs10793085TC0.460.478.59E-060.2894intergenicMIR548AL, KCNE311126,421,415rs59826922AG0.060.085.65E-060.7473intronic KIRREL3 1654,032,493rs11646260GA0.060.055.41E-060.0353intronic FTO 192,558,131rs78619038GA0.180.215.38E-060.1617intronic GNG7 1939,993,470rs55741253TA0.240.091.84E-060.0031exonic DLL3 2246,614,274rs1800206CG0.160.061.53E-060.4838exonic PPARA CHR chromosome, BP base pair position (GRCh37/hg19), A1 minor allele in NHGRI Family Cohort, A2 major allele in NHGRI Family Cohort, MAF1 minor allele frequency estimated from 1106 NHGRI Family Cohort founders, MAF2 minor allele frequency estimated from 208 NCR Cohort founders Meta-analysis results (P < 1 × 10−5) of the NHGRI family and NCR family cohorts CHR chromosome, BP base pair position (GRCh37/hg19), A1 minor allele in NHGRI Family Cohort, A2 major allele in NHGRI Family Cohort, MAF1 minor allele frequency estimated from 1106 NHGRI Family Cohort founders, MAF2 minor allele frequency estimated from 208 NCR Cohort founders Linkage results Two linkage regions on 2p25.2-p25.1 and 19p13.2-p13.11 reached a LOD score > 3.0 for the NHGRI Family Cohort (Figure S3), while 22q13.2-q13.31 reached the threshold for suggestive linkage (LOD > 2.0). For the NCR Family Cohort, while no significant linkage peaks were identified (at LOD > 3), there were suggestive linkage regions (at LOD > 2.0) at 3p26.3–26.2.2, 11p15.1 and 21q22.13–22.3.3 (Figure S4). For the meta-analysis, a linkage peak reaching the significant threshold LOD > 3.0 (P < 0.0001) was identified in a 4.9 Mb (± 1 Mb of chr19:13284603–16223287) region on 19p13.2–13.11.11 (Fig. 2), with five suggestive linkage peaks on 2p25.2-p25.1, 9p21-21.1, 11p15.1–14.2.2, 16p13.3-p13.2 and 22q13.2–13.31.31. We examined overlaps in these linkage signals with those reported in the prior literature (specifically 13 linkage regions with LOD > 3.0 in studies given in Table S3). We found that two of the six linkage regions (at LOD > 2.0) in our meta-analysis replicated previously reported linkage signals, with peaks on 2p25 and 16p13 (hypergeometric test P < 0.01). Fig. 2Plots of the a genome-wide linkage meta-analysis p-values (-log10(P)) of the NHGRI Family and NCR Family Cohorts and b for the significant region on chromosome 19 against the position of the variants where the horizontal blue lines correspond to LOD > 3.0 (~ P < 0.0001), while the horizontal red line corresponds to LOD > 4.0 (~ P < 0.00001). DTI diffusion tensor imaging, fMRI functional magnetic resonance imaging, AF L fractional anisotropy of left arcuate fasciculus, AST L fractional anisotropy of left anterior spinothalamic tract, CC fractional anisotropy of left corpus callosum, IFOF R fractional anisotropy of right inferior fronto-occipital fasciculus, ILF bilateral: fractional anisotropy of inferior longitudinal fasciculus, SLF bilateral: fractional anisotropy of superior longitudinal fasciculus, UF L fractional anisotropy of left uncinate fasciculus, Default mode—dorsal attention: connectivity between default mode and dorsal attention networks. Default mode -frontoparietal network: connectivity between default mode and frontoparietal control networks. Default mode—salience/ventral attention: connectivity between default mode and salience-ventral attention networks. Default mode—somatomotor network: connectivity between default mode and somatomotor network. P-values are shown only for nominal significant pairs (p< 0.05), with X indicating an FDR significant q-value (q < 0.05) Plots of the a genome-wide linkage meta-analysis p-values (-log10(P)) of the NHGRI Family and NCR Family Cohorts and b for the significant region on chromosome 19 against the position of the variants where the horizontal blue lines correspond to LOD > 3.0 (~ P < 0.0001), while the horizontal red line corresponds to LOD > 4.0 (~ P < 0.00001). DTI diffusion tensor imaging, fMRI functional magnetic resonance imaging, AF L fractional anisotropy of left arcuate fasciculus, AST L fractional anisotropy of left anterior spinothalamic tract, CC fractional anisotropy of left corpus callosum, IFOF R fractional anisotropy of right inferior fronto-occipital fasciculus, ILF bilateral: fractional anisotropy of inferior longitudinal fasciculus, SLF bilateral: fractional anisotropy of superior longitudinal fasciculus, UF L fractional anisotropy of left uncinate fasciculus, Default mode—dorsal attention: connectivity between default mode and dorsal attention networks. Default mode -frontoparietal network: connectivity between default mode and frontoparietal control networks. Default mode—salience/ventral attention: connectivity between default mode and salience-ventral attention networks. Default mode—somatomotor network: connectivity between default mode and somatomotor network. P-values are shown only for nominal significant pairs (p< 0.05), with X indicating an FDR significant q-value (q < 0.05) Overlap between linkage and FBAT signals From the NHGRI Family Cohort, the 19p13.2–13.11.11 significant linkage peak overlaps with rs55741253, exonic to DLL3 (P = 2.68 × 10− 8), while the meta-analysis suggestive linkage peak on 22q13.2-q13.31 encompasses rs1800206, exonic to PPARA, which had a P = 1.53 × 10− 6 in the FBAT meta-analysis. Comorbidity Of the 430 children with ADHD in the NHGRI Family Cohort, 184 (42.8%) also had oppositional defiant disorder (ODD), and only 21 with ODD did not have ADHD. We found considerable similarity between the FBAT results for ADHD (regardless of the presence or absence of comorbid ODD) and those for ADHD with ODD, and ADHD without this comorbidity. Specifically, of the 19 genome-wide significant variants in the analysis of ADHD, five were genome-wide significant in the FBAT analysis of ADHD only. In the FBAT of ADHD with comorbid ODD, 16 of the 19 SNPs from the primary analysis retained genome-wide significance, while an additional 3 variants reached genome-wide significance- details in Table S4. Convergence between family-based analyses and GWAS findings Using WebGestalt, we found that genes associated with ADHD in the NHGRI Family Cohort and in the meta-analysis enriched genes implicated in ADHD through GWAS (Demontis et al. 2023) (p-value = 0.013; FDR qvalue = 0.046). We also found that the polygenic risk score for ADHD, derived from independent data (Demontis et al. 2019), was associated with ADHD in both family cohorts. For the NCR Family Cohort, the maximal association was at the threshold of PT =1 (Odds Ratio of 1.46, 95% confidence interval of 1.18 to 1.80, p-value = 0.0005; FDR qvalue = 0.0036) and for the NHGRI Family Cohort, the maximum association was at the threshold of PT =0.1 (Odds Ratio of 1.18, 95% confidence interval of 1.05 to 1.34, p-value = 0.007; FDR qvalue = 0.07). Polygenic risk for autism spectrum disorder (ASD) was nominally associated (P < 0.05) with ADHD within the NCR Family Cohort only; otherwise, polygenic risk for other disorders was not tied to ADHD within either family cohort (Table S5). Gene based association and gene set enrichment analysis Gene-based association analyses using MAGMA on the NHGRI Family Cohort analyses (Figure S5) identified three exome-wide significant genes (CYP21A2, OR52I2 and PTGER2; P < 3.08 × 10− 6 (0.05/16192 genes). We submitted these three genes combined with the 31 genes implicated by an associated SNP GWAS signal in the NHGRI Family Cohort (P < 5 × 10− 8) to a gene-set enrichment analysis in the GENE2FUNC module in FUMA. There was significant enrichment of genes previously implicated by GWAS in autistic spectrum disorders or schizophrenia (Table 4). MAGMA did not identify any exome-wide genes associated with ADHD in the NCR cohort or the meta-analyses (Figure S6). Finally, we found enrichment by ADHD-associated genes in the NHGRI Family Cohort of two gene pathways: pattern recognition receptor activity (GO:0038187) and demethylase activity (GO:0032451) (Table S6). Table 434 genes for NHGRI family cohort: GWAS catalog reported genes for NHGRI family cohortGene set N n P-valueFDR q-valuesGenesAutism spectrum disorder or schizophrenia61268.25E-053.53E-02PPP1R11, MUC22, C6orf15, PSORS1C1, CDSN, CYP21A2Hematology traits2533.83E-066.95E-03C6orf15, PSORS1C1, CDSNLung cancer18747.83E-053.53E-02PITX2, MUC22, C6orf15, PSORS1C1White blood cell count19549.21E-053.53E-02PLCB1, C6orf15, PSORS1C1, CDSNEpstein-Barr virus immune response (EBNA-1)1229.73E-053.53E-02C6orf15, PSORS1C1N number of reported genes in gene set of GWAS catalog, n number of overlapped genes in our 34 genes with genes in gene set of GWAS catalog genes, Genes overlapped genes Table 434 genes for NHGRI family cohort: GWAS catalog reported genes for NHGRI family cohortGene set N n P-valueFDR q-valuesGenesAutism spectrum disorder or schizophrenia61268.25E-053.53E-02PPP1R11, MUC22, C6orf15, PSORS1C1, CDSN, CYP21A2Hematology traits2533.83E-066.95E-03C6orf15, PSORS1C1, CDSNLung cancer18747.83E-053.53E-02PITX2, MUC22, C6orf15, PSORS1C1White blood cell count19549.21E-053.53E-02PLCB1, C6orf15, PSORS1C1, CDSNEpstein-Barr virus immune response (EBNA-1)1229.73E-053.53E-02C6orf15, PSORS1C1N number of reported genes in gene set of GWAS catalog, n number of overlapped genes in our 34 genes with genes in gene set of GWAS catalog genes, Genes overlapped genes 34 genes for NHGRI family cohort: GWAS catalog reported genes for NHGRI family cohort N number of reported genes in gene set of GWAS catalog, n number of overlapped genes in our 34 genes with genes in gene set of GWAS catalog genes, Genes overlapped genes Overlap of ADHD genes and associated neural phenotypes For the NCR Family Cohort, we also identified genes associated with brain phenotypes, by first conducting FBAT analyses on these brain-based features. Using the GeneOverlap package, we found a significant overlapping between genes conferring ADHD risk within the NCR Family Cohort and genes associated with total white matter surface area (p-value: 0.006; q-value:0.019)- Fig. 3. There was also nominally significant genetic overlap between ADHD and one feature of the brain’s functional connectivity – connectivity between the default mode and somatomotor networks (p-value: 0.043). Fig. 3Gene overlap between ADHD with neuroimaging phenotypes for the NCR Family Cohort (P < 0.05) Gene overlap between ADHD with neuroimaging phenotypes for the NCR Family Cohort (P < 0.05) Discussion There are four main findings in this report on families affected by ADHD. First, nineteen SNPs were associated with ADHD at genome-wide levels of significance in a cohort of 1538 individuals from 359 nuclear families. When combined meta-analytically with a smaller cohort of 157 families (631 individuals), three of these SNPs remained significant at genome-wide levels of significance and twelve were suggestive association (at P < 1 × 10− 5). Of these, only one (rs77176301) demonstrated a consistent direction of effect across both cohorts. This may be due to the small size of the NCR cohort and low allele frequencies, which can produce uninformative Phet scores. When comparing these findings to a recent ADHD GWAS (Demontis et al. 2023), no overlap at the SNP or gene level was observed with the loci identified in our family-based analyses. This suggests that the signals identified here may represent independent associations and may reflect differences in study design, where variants enriched in families may be too rare to detect in case–control association studies. We also searched for a replication between our gene sets and rare variants recently reported as associated with ADHD (Demontis et al. 2025). There was functional convergence at the pathway level, but no gene-family replication between the gene sets. For example, the KCNE2, KCNE3 and KCNA2 discovered in our study are neuronal/ion channel signaling genes, as are CACNA1D and ANK2 identified by Demontis et al. (2025). Second, three regions segregating with ADHD in linkage analyses overlap significantly with previously identified ADHD linkage regions: 2p25.1 (LOD = 3.58) (Saviouk et al. 2011), 16p13.3-p13.2 (LOD = 4.2) (Smalley et al. 2002) and 19p13.2 (LOD = 1.72) (Arcos-Burgos et al. 2004). Third, we found a significant overlap between genes implicated in ADHD through our family-based analyses and those associated with ADHD in GWAS of unrelated cases and controls. Finally, using neuroimaging and genetic data obtained on the same individuals, we noted an overlap between the genes associated with ADHD and genes associated with some neuroanatomic and functional connectivity features. The family-based analyses identify potential new candidate genes, which, if confirmed, may point to genes that are more strongly tied to familial than sporadic ADHD. The meta-analysis identified genome-wide significant associations implicating TFRC, GSDME and TRIM31. TFRC plays a role in ferroptosis, by importing iron into the cells from the extracellular environment (Feng et al. 2020). GSDME belongs to the gasdermin family of genes and plays a role in pryoptosis (De Schutter et al. 2021; Wang et al. 2017), and has been found to harbor variants identified as pleiotropic loci associated with multiple psychiatric disorders including ADHD (Cross-Disorder Group of the Psychiatric Genomics Consortium. Electronic address and Cross-Disorder Group of the Psychiatric Genomics 2019) TRIM31 plays a role in the inflammatory response to infection (Rajsbaum, Garcia-Sastre, and Versteeg 2014) and was also found to be associated to intelligence in individuals ascertained for ADHD (Rizzi et al. 2011). The other gene emerging in the meta-analyses was GSDME (also known as DFNA5) which. Among the genes further implicated by the genome-wide significant associations in the larger NHGRI Family Cohort were PEX2, which has been found to have altered methylation at birth among those with childhood histories of ADHD (Walton et al. 2017); and NF1B, which was found to be haploinsufficient in 18 individuals with various neurodevelopmental phenotypes, including ADHD (Zhou et al. 2008; Schanze et al. 2018). This study complements the recent GWAS using unrelated cases and controls from the PGC in two ways. First, we find that polygenic risk for ADHD, determined from the independent data of the PGC, was associated with ADHD within both family cohorts. This is perhaps unsurprising given that most of those with ADHD within the PGC cohorts are likely to have family histories of the disorder (Rovira et al. 2020). Second, the genes associated with familial ADHD in our study significantly enriched genes implicated in ADHD through the PGC GWAS. We note that polygenic risk for other psychiatric disorders did not associate with ADHD in the larger NHGRI Family Cohort, possibly reflecting its stringent exclusion criteria. There was however a nominally significant association between ADHD within the NCR Family Cohort and polygenic risk for ASD. In gene set enrichment of broader biological categories (using GO tools), there was enrichment of genes pertaining to demethylase activity, which might point to epigenetic dysregulation as a mechanism of interest in ADHD (Cecil and Nigg 2022). Further functional annotation of the identified loci, including integration of eQTL data from resources such as GTEx, may help refine gene prioritization and improve the biological interpretation of these findings in future studies. We found an overlap between genes conferring risk for ADHD and those tied to variation in two neural features: total white surface area, and, at nominal levels of significance, a feature of functional connectivity feature that reflected interactions between the default mode network and a network involved in motor planning and action. The latter finding is of note as an influential model of ADHD posits the disorder to stem partly from dysregulated connectivity between the ‘task-oriented’ networks that support motor control and the default-mode network that is prominent during task-free, introspective processing (Sonuga-Barke and Castellanos 2007; Sutcubasi et al. 2020; Sudre et al. 2017, 2018). An analogy is that the online activity of the brain in motor planning and execution is repeatedly interrupted by the brain going ‘offline’, leading to symptoms characteristics of impulsivity and hyperactivity (Duffy et al. 2021, Rosenberg et al. 2016, Bozhilova et al. 2018). There are several limitations. Firstly, we consider only genotype array data and thus lacked the whole exome or whole genome sequences that are needed to capture rare coding and noncoding variants and some forms of structural variation that may be associated with ADHD. The cohorts were genotyped on different arrays, with less than 100,000 shared variants available for analysis prior to imputation. We limited the analysis of imputed data to variants with a MAF > 5%, since variants with lower frequencies have higher imputation inaccuracies, especially when dealing with family data (Stahl et al. 2021). While we were able to show similarity of associations for those who had ADHD either with or without comorbid ODD, other comorbid disorders were not prevalent enough to allow for a systematic exploration of their genetic associations within our families. Our cohorts also lacked detailed information across both cohorts regarding exposures to environmental factors, such as childhood adversity, thus not allowing to test for the possibility of a gene-environment correlation with our candidate genes. In conclusion, using a family-based approach, we report some novel genomic associations and linkage regions with familial ADHD and note overlaps with recent GWAS studies. In preliminary work, we point to changes in brain surface area and functional connectivity as candidates for the neural substrates underpinning these genomic findings. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1