A genomic study of trichotillomania and excoriation disorder in families Authors: Samantha R Greenspun, Isabella Milanes, Luis C Farhat, Sarah Abdallah, Diana Bok, Doris Chen, Wenzhong Liu, Enock Teefe, Michael H Bloch, Thomas V Fernandez, Emily Olfson Published: 2026 ABSTRACT AbstractTrichotillomania and excoriation disorder are obsessive-compulsive related disorders that are often subclassified together as body-focused repetitive behavior (BFRB) disorders. While previous research suggests shared genetic factors, the genetic architecture of these BFRBs remains incompletely understood. Probands with trichotillomania and/or excoriation disorder and both of their biological parents were recruited for an ongoing genetic study of parent-offspring trios with BFRBs. Genome-wide array data were generated in 110 families (334 individuals total) to investigate the role of both common single-nucleotide polymorphisms and rare copy-number variants (CNVs). Polygenic scores were calculated using summary statistics from genome-wide association studies of related psychiatric conditions, including obsessive-compulsive disorder (OCD), depression, anxiety, and attention-deficit/hyperactivity disorder. Using the polygenic transmission disequilibrium test (pTDT), we observed a significant over-transmission of the OCD polygenic score in probands of European ancestry from their parents (mean pTDT = 0.36, p = 0.01, n = 92), and a non-significant enrichment for the other conditions. Our results suggest that common variants associated with OCD may contribute to risk for BFRBs, consistent with their current classification as obsessive-compulsive related disorders. We also identified several rare CNVs in probands that overlapped genes intolerant to loss-of-function (LoF) mutations and those previously associated with neurodevelopmental disorders. The LoF-intolerant genes were enriched in biological processes relevant to synapse organization and neurodevelopment. This work provides new insight into the genetic underpinnings of these BFRB disorders, paving the way for larger genomic studies of these understudied conditions.Subject terms: Genomics, Psychiatric disorders Introduction Trichotillomania (TTM) and excoriation disorder (ED) are obsessive-compulsive related disorders (OCRDs) that impact 1–3% of the population [1–4]. These body-focused repetitive behavior (BFRB) disorders are characterized by frequent hair pulling or skin picking that results in hair loss or skin damage despite attempts to stop, leading to significant distress and/or functional impairment [5]. These disorders typically onset in early adolescence [6, 7] and are characterized by a waxing-and-waning clinical course [8–10]. Historically, TTM and ED were thought to be more common in females than males, though recent evidence suggests a more equal gender distribution in the general population [1–4]. TTM and ED frequently co-occur with other psychiatric conditions, including obsessive-compulsive disorder (OCD), depression, anxiety, and attention-deficit/hyperactivity disorder (ADHD) [2, 3, 11]. Currently, these conditions remain difficult to treat, with limited evidence-based treatments available [12–15]. This highlights a need for a greater understanding of the etiology of TTM and ED, with the hopes that advancements in that knowledge may guide the development of interventions. Previous family and twin studies suggest that genetic factors contribute to the development of BFRB disorders [16–20]. In a large twin registry study, Monzani et al. [21] demonstrate that the co-occurrence of OCRDs is best explained by two latent factors, the first of which is common to all OCRDs, while the second is specific to TTM and ED. Despite this role of genetic factors in the etiology of BFRB disorders, there is only one recent genome-wide analysis examining TTM [22] and none of ED. In this recent study, Halvorsen et al. [22] demonstrate a higher polygenic load for generalized psychiatric conditions in 101 individuals with TTM compared to controls and identify copy number variants (CNVs) previously associated with other neuropsychiatric disorders, highlighting the value of genomic investigations of BFRB disorders. Given evidence that similar genetic factors contribute risk to both TTM and ED [21], we sought to build on this past work by investigating the genetics of TTM and ED together using a parent-offspring trio design enabling the assessment of transmission of polygenic scores and leveraging summary statistics from a recently available genome-wide association study (GWAS) of OCD [23] to characterize if genetic factors associated with OCD contribute to these BFRB disorders. In this study, we collected and generated genome-wide array data in 110 families (334 individuals total), consisting of at least one proband with TTM and/or ED and both of their biological parents, making this the first parent-offspring trio genomic study of TTM and ED. This trio design offers advantages over case-control studies by controlling for potential bias in case-control matching, including biases related to population stratification and shared environmental factors [24, 25]. In these parent-offspring trios, we investigate the transmission of polygenic scores from parents to probands for several commonly co-occurring psychiatric conditions. Given prior evidence for shared genetic factors and clinical similarities across OCRDs [21], we hypothesized that probands with TTM and/or ED would have a significant over-transmission of the OCD polygenic score. Polygenic scores for major depressive disorder [26], anxiety disorders [27], and attention-deficit/hyperactivity disorder [28] were also examined to assess potential shared genetic risk between these conditions since they frequently co-occur with TTM and ED [2, 3, 11]. Given recent CNV findings by Halvorsen et al. [22] in TTM and previous CNV studies of OCD [29, 30], we also hypothesized that probands with TTM and/or ED may harbor rare CNVs that impact genes intolerant to loss-of-function (LoF-intolerant) mutations and neurodevelopmental risk genes. Overall, this study provides new insight into the genetics of two understudied BFRB disorders, TTM and ED, and lays the groundwork for future larger-scale genomic investigations. Methods Ethics approval and consent to participate Participants were enrolled in either an ongoing genetic study of BFRB disorders [11, 31] or a follow-up study of a pediatric TTM clinical trial [32]. Both studies were approved by the Institutional Review Board (IRB) at Yale School of Medicine (HIC 0301024156 and HIC 2000022993, respectively). All study procedures were conducted in accordance with the U.S. Federal Policy for the Protection of Human Subjects. Informed consent was obtained from all adult participants, and parental consent along with age-appropriate assent were obtained from minors. Participants and procedures This study involved the recruitment of parent-offspring trios from 110 families in which at least one proband reported a diagnosis of TTM and/or ED. One hundred of these families were recruited as part of an ongoing genetic study (previously described in [11, 31]) via the TLC Foundation for body-focused repetitive behaviors annual meetings (2018 and 2019) and online through advertisements on the TLC Foundation website and listserv, the study website (tabsstudy.org), as well as other support groups and organizations. The remaining 10 families were recruited as part of a follow-up study of a pediatric TTM clinical trial [32]. As previously described [11, 31], each participant was asked to complete a self-reported survey about their demographics, psychiatric and other diagnoses, family history, and symptoms (e.g., the Massachusetts General Hospital Hair-Pulling Scale was collected to assess TTM [33, 34], and the Skin-Picking Scale-Revised [35] was collected for ED). Genotyping and sample-level quality control Samples were genotyped using the Illumina Infinium Global Screening Array-24 v3.0 BeadChip and aligned to the GRCh38 (hg38) genome build at the Yale Keck Biotechnology Resource Laboratory. Quality control was performed using PLINK2 (https://www.cog-genomics.org/plink/2.0/, Latest Release, 01.29.2025). Individuals were excluded if they had genotype call rates below 98%, high or low levels of heterozygosity (3 standard deviations from the sample mean), or if discrepancies were detected between self-reported gender and sex chromosome data [36]. Relatedness was assessed using KING (https://www.kingrelatedness.com/, Latest Release, 07.28.2024), which uses genetic data to confirm reported relationships [37]. In the presence of admixture, KING accurately infers first-degree (parent-offspring, full siblings, dizygotic twins; kinship coefficient 0.1770–0.3540), second-degree (0.0884–0.1770), and third-degree (0.0442–0.0884) relationships. Any unexpected relationships within or between families were further investigated. Samples with discrepancies in relatedness were removed from the dataset. Overall, 101 probands and their biological parents (289 individuals total) passed these initial quality control (QC) steps (characteristics of 101 probands in Table 1; QC details are shown in Supplementary Fig. 1). Table 1Demographic and clinical characteristics of probands with trichotillomania and/or excoriation disorder.CharacteristicsProbands (n = 101)Sex, n (%) Male6 (6%) Female95 (94%)Age in years, mean (range)a22 (4–46)Diagnoses, n (%)a Trichotillomania44 (43%) Excoriation disorder21 (21%) Both BFRB disorders34 (34%)Co-occurring diagnoses, n (%)a Anxiety77 (76%) Depression42 (41%) Obsessive-compulsive disorder38 (38%) Attention-deficit/hyperactivity disorder22 (22%) Tourette and other tic disorders9 (9%) Autism spectrum disorder4 (4%) Intellectual disability3 (3%) Seizures1 (1%) Other psychiatric disorder15 (15%) Other neurological disorder5 (5%)Ancestry, n (%)b European95 (94%) Admixed American4 (4%) East Asian1 (1%) Central and South Asian1 (1%)Family structure, n (%)a Simplex41 (40%) Multiplex58 (57%)Characteristics of 101 probands with trichotillomania and/or excoriation disorder who passed initial quality control of genotyping data and are included in the genomic analyses.aTwo probands who completed consents and reported TTM and/or ED at the time of recruitment, but did not complete the surveys, so their age, diagnoses, and family structure are not included in this table.bAncestry was determined by principal components analysis. Demographic and clinical characteristics of probands with trichotillomania and/or excoriation disorder. Characteristics of 101 probands with trichotillomania and/or excoriation disorder who passed initial quality control of genotyping data and are included in the genomic analyses. aTwo probands who completed consents and reported TTM and/or ED at the time of recruitment, but did not complete the surveys, so their age, diagnoses, and family structure are not included in this table. bAncestry was determined by principal components analysis. Imputation and polygenic score analysis For polygenic score (PGS) analysis, we also restricted the dataset to single-nucleotide polymorphisms (SNPs) that passed additional QC filters. SNP-level filtering excluded variants due to missing genotype data (< 1%), variants that failed Hardy-Weinberg equilibrium testing (p < 1 × 10−6), and variants with minor allele frequencies (MAF) below 1% using PLINK2 [36]. After QC, 478,854 SNPs remained in the 101 parent-offspring trios for downstream analysis. Imputation was then performed on the QC data using the Sanger Imputation Service against the 1000 Genomes Phase 3 (version 5) reference panel and pre-phased using Eagle [38]. The 1000 Genomes Phase 3 panel was selected because it exists as a native GRCh38 (hg38) genome build and includes multi-ancestry representation [38]. Post-imputation, we retained 1,204,449 loci with an imputation quality score (R squared or estimated R²) of 0.3 or higher. We used GenoPred (opain.github.io/GenoPred) to calculate PGS using the PRS- Continuous Shrinkage (PRS-CS) method, which models linkage disequilibrium information and has been reported to outperform other commonly used PGS methods [39]. Scores were generated based on genome-wide association studies (GWAS) for OCD [23], major depressive disorder (MDD) [26], anxiety disorders (ANX) [27], and ADHD [28] under a phi of 1e-2 as recommended for highly polygenic traits (https://github.com/getian107/PRScs). All GWAS summary statistics were derived from European ancestry populations and were selected based on being the largest available datasets for each condition at the time of analysis. PGS were used to perform the polygenic transmission disequilibrium tests (pTDT, https://github.com/ypaialex/ptdt) in the 92 complete parent-offspring trios of European ancestry (Supplementary Fig. 2) to assess whether probands inherited a greater polygenic burden for psychiatric disorders than expected by chance. pTDT is calculated by subtracting the mean parental PGS from the proband’s PGS and standardizing this difference by the standard deviation of parental scores [40]. A positive pTDT score suggests that the affected proband inherited a greater genetic risk burden than expected by chance from their parents. CNV calling and quality control CNVs were called in all BFRB trios that passed general sample-level QC metrics (Supplementary Fig. 1) with Log R ratio (LRR) and B allele frequency (BAF) data with reference alleles relative to hg38. The calling procedure used PennCNV (http://penncnv.openbioinformatics.org) and cnvPartition (v3.2.1) to generate separate CNV call sets for each sample. CNV loci were validated based on the intersection of these two different call sets (50% overlap), and CNV calls present in both were retained. To ensure high-confidence CNV calls, we excluded poor-quality samples likely to contain spurious CNV calls. Quality control metrics were computed using the PennCNV filter_cnv.pl command with thresholds defined according to PennCNV-recommended guidelines and previously published CNV studies, including Halvorsen et al. [29] and McGrath et al. [30]. Outlier detection focused on four sample-level metrics: Log R Ratio standard deviation (LRR_SD), absolute waviness factor (|WF|), BAF drift, and number of CNV calls (numcnv). Samples exceeding any of the following thresholds were removed: LRR_SD > 0.3, |WF | > 0.05, BAF drift > 0.01, or numcnv ≥ 20 [29, 30]. This resulted in a further exclusion of 15 trios from the initial 101 trios that passed QC, with the final set for CNV analysis comprising 86 trios (Supplementary Fig. 1). After applying these filters, all retained samples met the specified quality control criteria. Additional details on PennCNV and cnvPartition are in the Supplemental Methods. CNV filtering and annotation CNV calls underwent several filtering steps to generate a final set of rare CNVs (MAF < 1%) for downstream analyses, as done in previous studies [29, 30]. First, we removed CNVs that overlapped genomic regions (defined as > 50% overlap), known to harbor spurious CNV calls: immunoglobulin regions, segmental duplications, or loci within 500 kb of telomeres or centromeres [41]. We also removed CNVs where 50% of bases overlapped loci from Repeatmasker (version 4.0.5, http://www.repeatmasker.org) or “non-defined” (i.e., polyN) portions of the hg38 reference panel. CNVs that were adjacent were merged into one single call if they overlapped by ≥ 30% of the total base-pair length. Calls were then subjected to a series of CNV size and frequency filters. Specifically, calls were required to be greater than 30 kb in size, found in less than 1% of the population in both gnomAD (version 4.1.0) structural variants global and European datasets (covering over 14,000 individuals), and in our CNV dataset. CNVs were also required to have a confidence level greater than 50, as done previously [42, 43]. We also examined rates of CNV calls that were both rare (MAF < 1%) and large (≥ 500 kb). Classification of CNVs as either de novo or transmitted was based on parental CNVs and was confirmed by visualizing both the offspring’s and parents’ CNV profiles. See the Supplemental Methods section for a description of the full procedure. Rare CNVs that passed filtering criteria were annotated to genes using the UCSC hg38 database. CNV calls were categorized as deletions if copy number ≤ 1 or duplications if copy number ≥ 3. Genes overlapping with CNVs were then assessed for loss-of-function (LoF) intolerance, defined as a pLI score ≥ 0.9 in gnomAD (version 4.1.0). These genes were also checked to see if they have been reported in other neuropsychiatric and neurodevelopmental conditions. This included CNVs previously associated with neurodevelopmental disorders defined by Kendall et al. [44], and CNVs predicted to disrupt neurodevelopmental risk genes defined by an FDR ≤ 0.05 in Fu et al. [45]. These categories were chosen to match a recent large, case-control CNV study in OCD [29] to enable comparison. Statistical analyses Statistical significance of pTDT deviation from zero was determined using a two-sided, one-sample t-test, and variation in pTDT was quantified by standard error as done in previous studies [24, 40, 46] (Fig. 1A and B). Normal distribution of pTDT deviation scores was evaluated using density plots with normal distribution overlays and QQ-plots, which both indicated no major departures from normality in our dataset. For analyses comparing PGS or pTDT scores between sub-phenotype groups (Fig. 1B, Supplementary Fig. 3 and Supplementary Fig. 4A), two-sided, two-sample t-tests or the analysis of variance (ANOVA) test were used as appropriate. OCD PGS of simplex and multiplex probands and their parents were standardized to z-scores to allow comparability across groups (Supplementary Fig. 4A). Differences in rates of CNVs between simplex and multiplex probands (Supplementary Fig. 4B) and between proband and parent groups (Table 2 and Supplementary Table 1) were compared using Fisher’s exact test. Transmission of polygenic scores for psychiatric conditions in probands of European Ancestry with trichotillomania and/or excoriation disorder. A Transmission of polygenic scores for several related psychiatric conditions in probands with body-focused repetitive behavior disorders (BFRBs). The polygenic transmission disequilibrium test (pTDT) compares the mean deviation of proband polygenic scores to the mid-parent distribution. A significantly increased transmission of the polygenic score for obsessive-compulsive disorder (OCD) was observed in the probands with BFRBs (mean pTDT = 0.36, p = 0.01). Transmission of polygenic scores were not statistically significant for other neuropsychiatric conditions (MDD pTDT = 0.15, p = 0.17; ANX pTDT = 0.08, p = 0.44; ADHD pTDT = 0.01, p = 0.94). Error bars are ± 1 standard error. ∗p < 0.05 in a two-sided, one-sample t-test. B Sensitivity analysis examining transmission of the OCD polygenic score in BFRB probands with or without a self-reported co-occurring OCD diagnosis. Probands with BFRBs who have an OCD diagnosis and those who do not show a mean pTDT in the positive direction, but it was not statistically significant in either group (mean pTDTNo co-occurring OCD = 0.35, p = 0.06, n = 57; mean pTDTCo-occurring OCD = 0.39, p = 0.05, n = 35). The difference between the groups was not statistically significant (p = 0.88). OCD obsessive-compulsive disorder, MDD major depressive disorder, ANX anxiety disorders, ADHD attention-deficient/hyperactivity disorder. Table 2Rates of rare copy number variants in probands with trichotillomania and/or excoriation disorder.Number of CNVsRate of CNVs per probandProportion of probands with ≥ 1 CNVRare CNVs, >30 kb All CNVs620.720.36 Deletion270.310.21 Duplication350.400.24Large rare CNVs, >500 kb All CNVs40.050.05 Deletion10.010.01 Duplication30.030.03Rare CNVs, >30 kb, overlapping genes with pLI ≥ 0.9 All CNVs80.10.1 Deletion50.060.06 Duplication30.030.03Rare CNVs, >30 kb, overlap with neurodevelopmental CNVs or risk genes All CNVs70.080.07 Deletion30.030.03 Duplication40.050.03Rates of different categories of rare copy number variants (CNVs) identified by both PennCNV and cnvPartition in 86 probands with trichotillomania and/or excoriation disorder who passed quality control steps for CNV analyses. Rare CNVs were defined as < 1% in gnomAD (version 4.1.0) structural variants (global and European datasets) and in our CNV dataset. Large CNVs were defined as > 500 kb. Rare copy number variants (CNVs) impacting genes with high loss-of-function intolerance were defined based on probability loss-of-function intolerance (pLI) scores ≥ 0.9. CNVs previously associated with neurodevelopmental disorders were defined by Kendall et al. [44], and CNVs harboring neurodevelopmental risk genes with a false discovery rate (FDR) < 0.05 in Fu et al. [45]. kb kilobases.Note: rate of CNVs per proband = total number of detected CNVs / the number of probands who passed quality control (n = 86); proportion of probands with ≥ 1 CNV Proportion = number of unique probands with ≥ 1 CNV/ 86. Rates of rare copy number variants in probands with trichotillomania and/or excoriation disorder. Rates of different categories of rare copy number variants (CNVs) identified by both PennCNV and cnvPartition in 86 probands with trichotillomania and/or excoriation disorder who passed quality control steps for CNV analyses. Rare CNVs were defined as < 1% in gnomAD (version 4.1.0) structural variants (global and European datasets) and in our CNV dataset. Large CNVs were defined as > 500 kb. Rare copy number variants (CNVs) impacting genes with high loss-of-function intolerance were defined based on probability loss-of-function intolerance (pLI) scores ≥ 0.9. CNVs previously associated with neurodevelopmental disorders were defined by Kendall et al. [44], and CNVs harboring neurodevelopmental risk genes with a false discovery rate (FDR) < 0.05 in Fu et al. [45]. kb kilobases. Note: rate of CNVs per proband = total number of detected CNVs / the number of probands who passed quality control (n = 86); proportion of probands with ≥ 1 CNV Proportion = number of unique probands with ≥ 1 CNV/ 86. Exploratory gene-ontology analysis We used ConsensusPathDB (http://cpdb.molgen.mpg.de/, Latest Release 35, 05.06.2021) to evaluate whether the set of LoF-intolerant genes that were predicted to be disrupted by a rare CNV (Tables 2 and 3) was significantly over-represented in 31 public gene ontology sets. Supplementary Table 3 provides the p-values (< 0.01) and q-values (< 0.05) for all enriched gene ontology sets. Additional details on how this data was generated can be found in the Supplementary Methods. Table 3Rare copy number variants identified in probands with BFRBs that overlap with LoF-intolerant genes, neurodevelopmental CNVs/loci, and neurodevelopmental risk genes.IDSexBFRB diagnosisCo-occurring conditionsGenomic coordinates (hg38)CNV classSize, bpKnown CNV/LocusGene(s) of interest impacted (pLI score if ≥ 0.9)Other known genes impactedBFRB9003.p1FemaleEDOCD, ANX2:240684041- 240762707DUP78,667-KIF1A (1)AQP12A8:73040513- 73159092DUP118,580-TERF1 (0.9)SBSPONBFRB9021.p1FemaleTTMMDD, ANX2:50652887- 50725225DEL72,339NRXN1 delNRXN1 (1)MIR8485BFRB9023.p1FemaleTTMOCD, ANX7:73308984- 73391588DUP82,605WBS dupFKBP6, TRIM5FKBP6, TRIM5,RNU6-1080P, RPL7AP77BFRB9047.p1FemaleTTMOCD, MDD, ANX5:37000986-37046181DEL45,196-NIPBL (1)-10:87894024-87925562DEL31,53910q22-q23PTEN-BFRB9059.p1FemaleTTMANX9:101144710-101244151DEL99,442-PLPPR1 (0.94)-BFRB9061.p1FemaleTTMADHD, ANX15:24140203-24249920DEL109,718PWS/AS delPWRN1PWRN2BFRB9072.p1FemaleEDOther BFRB, ANX8:8243957- 8465532DUP221,5768p23.1 dupPRAG1FAM86B3P, LINC02949, LINC02950BFRB9082.p1FemaleTTMMDD, ANX14:19202105- 19948340DUP746,236-POTEG (1)ARHGAP42P4, BMS1P17, CDRT15P13, DUXAP10, GRAMD4P3, LINC01297, MED15P6, NBEAP6, NF1P11, NF1P4, OR11H2, OR11K2P, OR4H12P, OR4K1, OR4K2, OR4K3, OR4K4P, OR4K5, OR4K6P, OR4M1, OR4N1P, OR4N2, OR4Q3, RNU6-1268P, TOMM40P1BFRB9083.p1FemaleTTM, EDOther BFRB, TD, MDD, ANX12:7853028-7970710DEL117,683-SLC2A14 (1)Metazoa_SRP, NANOGP1, SLC2A3BFRB9094.p1FemaleEDMDD, ANX20:42605461-42650964DEL45,504-PTPRT (1)-Rare copy number variants (CNVs) found in 10 probands with BFRBs that overlap loss-of-function (LoF) intolerant genes (pLI ≥ 0.9), known neurodevelopmental CNVs, and neurodevelopmental risk genes (indicated by boldness). CNVs that overlap regions of interest previously associated with neurodevelopmental disorders were defined by Kendall et al. [44], and CNVs harboring neurodevelopmental risk genes with a false discovery rate (FDR) ≤ 0.05 in Fu et al. [45].TTM trichotillomania, ED excoriation disorder, OCD obsessive-compulsive disorder, TD tic disorder, ANX anxiety disorder, MDD major depressive disorder, ADHD attention-deficient/hyperactivity disorder, DEL deletion, DUP duplication, WBS Williams-Beuren syndrome, PWS/AS Prader-Willi Syndrome/Angelman Syndrome. Rare copy number variants identified in probands with BFRBs that overlap with LoF-intolerant genes, neurodevelopmental CNVs/loci, and neurodevelopmental risk genes. Rare copy number variants (CNVs) found in 10 probands with BFRBs that overlap loss-of-function (LoF) intolerant genes (pLI ≥ 0.9), known neurodevelopmental CNVs, and neurodevelopmental risk genes (indicated by boldness). CNVs that overlap regions of interest previously associated with neurodevelopmental disorders were defined by Kendall et al. [44], and CNVs harboring neurodevelopmental risk genes with a false discovery rate (FDR) ≤ 0.05 in Fu et al. [45]. TTM trichotillomania, ED excoriation disorder, OCD obsessive-compulsive disorder, TD tic disorder, ANX anxiety disorder, MDD major depressive disorder, ADHD attention-deficient/hyperactivity disorder, DEL deletion, DUP duplication, WBS Williams-Beuren syndrome, PWS/AS Prader-Willi Syndrome/Angelman Syndrome. Results Demographic and clinical characteristics After quality control, we examined 101 probands with a diagnosis of TTM, ED, or both conditions and their biological parents (289 individuals total, including 87 parent-offspring trios and 7 quads with two affected probands). The mean age at assessment for probands was 22 years, and the cohort was predominantly female (94%) (Table 1). Based on the principal component analysis, most probands were of European ancestry (n = 95), with smaller representation from Admixed American ancestry (n = 4), East Asian ancestry (n = 1), and Central and South Asian ancestry (n = 1) (Table 1). Trios were classified as simplex when the proband had no first-degree relatives with a history of trichotillomania or excoriation disorder (n = 41), and as multiplex when at least one first-degree relative had a history of a BFRB (n = 58) (Table 1). Notably, almost all multiplex families (97%, 56/58) included at least one affected parent. Polygenic transmission for psychiatric conditions Results demonstrate a significant over-transmission of the OCD polygenic score in probands with BFRBs from their parents (mean pTDT = 0.36, p = 0.01) (Fig. 1A). For depression, anxiety, and ADHD, transmission of polygenic scores from parents to probands with BFRBs was not statistically enriched, though in the positive direction (Fig. 1A). We performed a sensitivity stratified analysis of probands who did and did not self-report a co-occurring OCD diagnosis (Fig. 1B), demonstrating a positive, non-significant transmission of OCD polygenic score in both groups (mean pTDTNo co-occurring OCD = 0.35, p = 0.06, n = 57; mean pTDTCo-occurring OCD = 0.39, p = 0.05, n = 35) with no statistically significant difference between these two groups (p = 0.88). To further understand patterns of OCD polygenic score transmission in these families with BFRBs, we also conducted post-hoc exploratory secondary analyses to investigate the impact of BFRB disorder diagnosis (TTM, ED, vs both) and family structure (simplex vs multiplex) (Supplementary Fig. 3A, Supplementary Fig. 3B, and Supplementary Fig. 4A). For BFRB diagnosis, the mean pTDT for the OCD polygenic score was in the positive direction for all three BFRB disorder diagnosis groups, but was only statistically significant for those with a singular diagnosis of TTM (mean pTDT = 0.56, p = 0.01, n = 36) (Supplementary Fig. 3A). Probands from simplex families showed a significant over-transmission of the OCD polygenic score (mean pTDT = 0.67, p = 0.001, n = 37), but not those from multiplex families (mean pTDT = 0.17, p = 0.34, n = 55) (Supplementary Fig. 3B). Consistently, simplex probands had a significantly higher OCD polygenic score compared to their parents (p = 0.009), while this difference was smaller and non-significant between multiplex probands and their parents (p = 0.49) (Supplementary Fig. 4A). We did not detect significant differences in the OCD polygenic score between simplex and multiplex probands (p = 0.42), or between parents from simplex versus multiplex families (p = 0.20). Copy-number variant analysis We assessed rates of rare (MAF < 1%) CNVs in probands with TTM and/or ED, showing an overall rate of 0.72 rare CNVs per proband, with a deletion rate of 0.31 and a duplication rate of 0.40 (Table 2). After visualization of parents, we did not detect any rare de novo CNVs. As a result, our downstream analyses focused on rare transmitted CNVs. In addition to rare CNVs, we also examined the rate of large (length ≥ 500 kb), rare CNVs in BFRB probands, with a deletion rate of 0.01 and a duplication rate of 0.03. Using the set of genes that overlapped rare CNVs identified in probands, we investigated whether these genes showed high intolerance to loss-of-function (LoF) variation or were previously associated with neurodevelopmental risk. Eight rare CNVs overlapped genes with high LoF intolerance scores (pLI ≥ 0.9) (Table 3). In addition, we identified five CNVs previously associated with regions of interest and/or loci implicated in neurodevelopmental disorders by Kendall et al. [44], including deletions in NRXN1,10q22-10q23, and 15q11.2-q13 (associated with Prader-Willi syndrome and Angelman syndrome), and duplications in 7q11.23 (associated with Williams–Beuren syndrome) and in 8p23.1 (Table 3 and Supplementary Table 2.1). We further mapped these CNVs to the specific genes they overlapped that have been associated with the neurodevelopmental disorder (NRXN1, FKBP6, TRIM50, PTEN, and PWRN1) [44]. Four rare CNVs overlapped high-confidence risk genes associated (FDR < 0.05) with neurodevelopmental disorders (NDDs) in Fu et al. [45], including KIF1A, NRXN1, NIPBL, and PTEN (Table 3 and Supplementary Table 2.2). The majority of the rare CNVs impacting LoF-intolerant or neurodevelopmental genes occurred in probands from multiplex families (8 of 12), but the difference in rates of multiplex and simplex probands with risk CNVs was not statistically significant (Supplementary Fig. 4B). Additionally, probands had a significantly higher rate of rare CNVs (>30 kb) than parents (p = 0.04) (proband CNV rates in Table 2; parental CNV rates in Supplementary Table 1). Rates of rare large CNVs (>500 kb), CNVs overlapping LoF-intolerant genes, and CNVs overlapping neurodevelopmental genes appeared higher in probands than parents, but these differences were not statistically significant. Visual representations of these individual CNVs and neurodevelopmental FDR results are available in the supplement (Supplementary Fig. 5 and Supplementary Table 2.2, respectively). Characteristics of individuals with these rare CNVs impacting LoF-intolerant and neurodevelopmental genes are detailed in Table 3, and it is notable that none of these probands self-reported a history of autism or intellectual disability. Exploratory gene ontology analysis Using the subset of LoF-intolerant genes disrupted by rare CNVs in probands with BFRBs, we conducted exploratory analyses to assess the enrichment of gene-ontology terms. Several gene ontology categories were significantly enriched among these eight genes, and the top enriched gene-ontology set is “regulation of synapse organization” (q-value = 0.0009) (Supplementary Table 3). Discussion In this first parent-offspring trio genomic study of trichotillomania and excoriation disorder, our findings highlight the role of common SNPs and rare CNVs in these body-focused repetitive behavior disorders, advancing our understanding of these understudied conditions. Our results show that probands with TTM and/or ED inherit a higher polygenic score for OCD from their parents than expected by chance (Fig. 1A). This significant over-transmission of the OCD polygenic score was observed in our primary analysis of all BFRB probands of European ancestry. Our sensitivity analysis comparing BFRB probands with and without a co-occurring OCD diagnosis shows a similarly non-significant positive transmission in both groups and no significant difference between groups, highlighting that the increased transmission of the OCD polygenic score in BFRB probands may not be fully explained by the high co-occurrence of OCD with BFRBs (Fig. 1B). In exploratory analyses, we also found a significantly elevated transmission of the OCD polygenic score in probands from simplex families. The magnitude of the over-transmission of the OCD polygenic score was smaller and non-significant in multiplex families where a first-degree relative had TTM and/or ED (Supplementary Fig. 3). Similarly, simplex probands had a significantly higher OCD polygenic score than their parents, while the proband-parent difference was smaller and nonsignificant in multiplex families (Supplementary Fig. 4A). There was no significant difference in the OCD polygenic score between simplex and multiplex probands, or between their parents. This suggests that inheriting an elevated polygenic score for OCD may predispose an individual to developing a BFRB, which may be especially relevant if they do not have a family history of BFRBs, though the difference between multiplex and simplex families was not statistically significant. Given the limited sample size of this study, replication of these findings in larger samples is needed. Overall, these findings highlight potentially shared genetic factors between OCD and BFRBs. This is consistent with TTM and ED being classified as OCRDs in DSM-5 [5], the high comorbidity between BFRBs and OCD [2, 3, 11], as well as family and twin studies that have long supported the shared liability between OCD and BFRBs [17, 18, 21, 47]. Our results also extend recent findings from Halvorsen et al. [22], who demonstrated a higher polygenic load for generalized psychiatric conditions in 101 cases with TTM compared to controls. Collectively, our findings along with this past work highlight that large-scale genome-wide studies of TTM and ED may advance our understanding of the role of common genetic variants in these understudied disorders in a similar fashion to recent studies in OCD and other neuropsychiatric conditions. Our study also examined if there was an over-transmission of polygenic scores for depression, anxiety, and ADHD in probands with TTM and/or ED, as these conditions commonly co-occur in our sample and in other studies [2, 3, 11]. Given this co-occurrence, we hypothesized that there may be an overlap in genetic factors. Although we did not observe statistically significant enrichment, all associations were in the positive direction (Fig. 1A), consistent with findings by Halvorsen et al. [22] who reported no differences in polygenic scores for depression and ADHD in trichotillomania cases compared to matched controls. Our lack of statistical enrichment may be due in part to our limited sample size, and further genetic studies are needed to further characterize shared genetic factors between BFRB disorders and these commonly co-occurring conditions. In this study, we also identified 62 rare CNVs (MAF < 0.01, length ≥ 30 kb) in probands with TTM and/or ED, with a rare CNV deletion rate of 0.31 and a duplication rate of 0.40 (Table 2). These rates are of a similar magnitude to previously reported rates of rare CNVs identified using similar methods in other studies of TTM and OCD [22, 29, 30]. Among 101 individuals with TTM, Halvorsen et al. [22] identified a rare CNV (length ≥ 100 kb) deletion rate of 0.27 and a duplication rate of 0.29. In a rare CNV study (length ≥ 30 kb) of OCD (2248 cases, 3608 unaffected controls), Halvorsen et al. [29] report a rare deletion rate of 0.27 and a rare duplication rate of 0.32 in OCD cases compared to a deletion rate of 0.25 and a duplication rate of 0.31 in unaffected controls. We also examined the subset of large, rare CNVs (length ≥ 500 kb) in BFRB probands, with a deletion rate of 0.01 and a duplication rate of 0.03, which are on a similar order of magnitude to rates previously reported by McGrath et al. [30] for OCD and Tourette’s disorder. In addition, probands had higher rates of rare CNVs than parents. Overall, these findings suggest that probands with BFRBs have rare CNV rates comparable to those reported in OCRDs, and potentially higher than rates previously observed in controls and in their parents. This underscores that larger-scale investigations of rare CNVs in BFRBs may provide additional insight into the genetic underpinnings of these disorders in a similar manner to previous studies of related conditions. After visualization, no reliable rare de novo CNVs were detected in the 86 probands with BFRBs. Previous genome-wide array studies have identified de novo CNVs in 1.4% of OCD cases and 2.9% of Tourette’s disorder cases [30, 48], with a higher rate identified in a whole-exome DNA sequencing analysis of pediatric OCD [49], which can detect smaller CNVs than array-based analyses. Our BFRB cohort included several multiplex families, which may also contribute to the lower de novo CNV yield than past studies [48, 50]. In this study, we identified 8 rare CNVs that overlapped loss-of-function (LoF)-intolerant genes (pLI ≥ 0.9) (Table 2). Previous studies suggest an enrichment of rare CNVs impacting LoF-intolerant genes in OCD [29, 30], and Halvorsen et al. [22] observed a nonsignificant difference in constrained CNVs between individuals with TTM compared to controls. Based on this past work, we performed exploratory gene ontology analysis of the 8 LoF-intolerant genes impacted by rare CNVs and found significant enrichment in several biological processes, such as synaptic organization, activity, and structure (Supplementary Table 3). This aligns with prior human and animal studies that implicate synaptic dysfunction in the pathophysiology of BFRBs [18, 51, 52], supporting the neurodevelopmental underpinnings of these disorders. Finally, we identified five rare CNVs overlapping genomic loci previously implicated in neurodevelopmental disorders defined by Kendall et al. [44] (Table 3, Supplementary Table 2.1), and two additional CNVs overlapping risk genes for neurodevelopmental disorders defined by an FDR < 0.05 in Fu et al. [45] (Table 3, Supplementary Table 2.2). Of note, none of the probands with these CNVs self-reported a history of intellectual disability or autism. One proband with TTM was identified to have a rare maternally inherited deletion in NRXN1, a well-established risk gene for neurodevelopmental disorders [44, 45]. Interestingly, Halvorsen et al. [22] also identified a different NRXN1 deletion (chr2:50790067-50956421, hg38) in an individual with TTM in a cohort collected independently. Another proband with TTM was identified to have a paternally inherited CNV associated with Williams-Beuren syndrome [44], overlapping FKBP6 and TRIM50 [53, 54]. A separate proband had two rare deletions overlapping NIPBL and PTEN, both of which are associated with neurodevelopmental disorders [44, 45] (Table 3 and Supplementary Tables 2.2). We also detected a CNV deletion in PWRN1 in a proband with TTM, which is associated with Prader-Willi and Angelman Syndromes [44]. In a proband with ED, we identified a CNV duplication that overlaps a region associated with 8p23.1 duplication syndrome, which is linked to learning disabilities and developmental delay [44]. Finally, one proband with ED had a duplication in KIF1A, which is associated with various neurological disorders [45, 55], though the implications of a duplication in this gene are less well understood. Overall, these findings suggest that there may be shared genetic factors between BFRBs and neurodevelopmental conditions [56]. Several limitations of this study should be considered. First, ideally, control trios would be genotyped alongside cases and assessed for BFRBs to enable a more accurate comparison of burden differences in polygenic score and CNVs. In this study, we prioritized genotyping affected parent-offspring trios and leveraged the pTDT test to compare polygenic scores in probands to their parents, which is a statistically well-powered approach to test if affected probands inherit an increased polygenic load from their parents. For rare CNVs, we compared rates in probands to those previously reported in large-scale studies of related disorders that used similar analysis methods, but future work in larger samples that simultaneously assess controls is needed to confirm the enrichment of rare CNVs in BFRBs. Second, our study examined polygenic scores for disorders commonly co-occurring with BFRBs to better understand the genomic architecture of these understudied disorders. An important limitation in the field is that there are currently no large-scale GWAS of TTM or ED, which would enable a more direct understanding of the role of common genetic variants in families impacted by BFRBs. Third, our sample had limited sex (94% female) and ancestry (95% European ancestry) diversity, and future efforts should focus on enrolling unrepresented groups to ensure our findings are generalizable. Fourth, diagnoses were based on self-report measures rather than semi-structured clinical interviews. Finally, our analysis focused on common variants and rare CNVs that can be assessed by genome-wide array data, but it is likely that other variants, including rare gene-damaging single-nucleotide and indels, contribute to BFRB risk. Future large-scale DNA sequencing studies may provide further insight into the full spectrum of genetic variation that contributes to BFRB risk. In conclusion, our study offers new insights into the genetic architecture of trichotillomania and excoriation disorder. For the first time, we reported a significant over-transmission of the polygenic score for OCD in probands with these body-focused repetitive behavior disorders. We also identified several rare CNVs that overlap LoF-intolerant and neurodevelopmental risk genes. These findings represent an important step forward toward understanding the genomic underpinnings of BFRBs. Future efforts leveraging larger BFRB cohorts may continue to deepen our understanding of trichotillomania and excoriation disorder genomics and pathophysiology. Supplementary information Supplementary Information Acknowledgements We would like to thank the families who contributed to this research study, as well as the TLC Foundation for BFRBs for their support of our recruitment efforts. We would also like to acknowledge research assistants Kayla Delapenha and Jessica Levine, who contributed to the recruitment and collection of data. Author contributions EO, MHB, and TVF designed the research. SRG, IM, LCF, SA, DB, DC, WL, ET, MHB, TVF, and EO performed the research. SRG analyzed the data. EO supervised the data analysis. SRG and EO drafted the manuscript. All authors critically reviewed the paper. Funding E.O. reports support for this research from the National Institutes of Mental Health (NIMH) of the National Institutes of Health (NIH) (grant number K08MH128665), the American Academy of Child and Adolescent Psychiatry, and the Alan B. Slifka Foundation Riva Ariella Ritvo endowment. T.V.F reports support for this research from the NIMH of the NIH (grant number R01MH114927). S.R.G. reports support from the National Center for Advancing Translational Science of the NIH (grant number TL1TR001864). I.M. reports support from the National Institute of Neurological Disorders and Stroke of the NIH (grant number 5T32NS041228-24). S.B.A. reports support from the NIMH of the NIH (grant numbers T32MH18268 and R25MH077823). The other authors report no relevant funding. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. Data availability Genotype microarrays and associated phenotype data from probands and their parents who provided consent for sharing are available through controlled access via the Database of Genotypes and Phenotypes (dbGaP) under accession phs004005.v1.p1. Code availability Raw genotype array data were processed using GenomeStudio’s PLINK Report Plugin (v2.14). Quality control, including relatedness and sex chromosome checks, was performed using PLINK v1.9 (build June 2021) and PLINK 2.0 (build 24 Oct 2022). Code used in this work to prepare and operate the imputation was provided under the Sanger Imputation pipeline (https://imputation.sanger.ac.uk/?about=1#pipeline). Polygenic scoring, as well as principal component analysis, was performed using Genopred (opain.github.io/GenoPred). Copy number variants were called using PennCNV (http://penncnv.openbioinformatics.org) and Illumina’s cnvPartition (v3.2.1), with filtering and quality control performed using code adapted from the PennCNV repository (http://penncnv.openbioinformatics.org). Competing interests The authors declare that there are no conflicts of interest. In the past three years, E.O. has received research support or grants from the National Institutes of Health, The Hartwell Foundation, Tourette Association of America, International OCD Foundation, Misophonia Research Fund, Yale Center for Clinical Investigation, and Yale Child Study Center. She serves as chair of the Early Career Investigator Program Committee and on the Board of Directors for the International Society of Psychiatric Genetics (unpaid). She is also a member of the Research Committee for American Academy of Child & Adolescent Psychiatry (unpaid). She is a sub-investigator on a clinical trial with support from Neurocrine Biosciences (unpaid). T.V.F. has received research support or grants from the National Institutes of Health, Misophonia Research Fund, New Venture Fund, and Yale Child Study Center. He received an honorarium for participation in the 2025 Pediatric Psychopharmacology Update Institute by the American Academy of Child & Adolescent Psychiatry. He is paid for expert testimony and consultation by DLA Piper LLC. M.H.B. reports grant or research support from SciSparc, Emalex Biosciences, Janssen Pharmaceuticals, Neurocrine Biosciences, and the National Institutes of Health. He has also received royalties from Wolters Kluwer for Lewis’s Child and Adolescent Psychiatry: A Comprehensive Textbook, Fifth Edition, and moonlighting pay from the Veterans’ Administration. None of these relationships is of direct relevance to the work described here. The other authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04164-6. References Associated Data Data Availability Statement Genotype microarrays and associated phenotype data from probands and their parents who provided consent for sharing are available through controlled access via the Database of Genotypes and Phenotypes (dbGaP) under accession phs004005.v1.p1. Raw genotype array data were processed using GenomeStudio’s PLINK Report Plugin (v2.14). Quality control, including relatedness and sex chromosome checks, was performed using PLINK v1.9 (build June 2021) and PLINK 2.0 (build 24 Oct 2022). Code used in this work to prepare and operate the imputation was provided under the Sanger Imputation pipeline (https://imputation.sanger.ac.uk/?about=1#pipeline). Polygenic scoring, as well as principal component analysis, was performed using Genopred (opain.github.io/GenoPred). Copy number variants were called using PennCNV (http://penncnv.openbioinformatics.org) and Illumina’s cnvPartition (v3.2.1), with filtering and quality control performed using code adapted from the PennCNV repository (http://penncnv.openbioinformatics.org).