Genome-wide association analyses identify distinct genetic architectures for early-onset and late-onset depression

Like other complex disorders such as type 2 diabetes¹ and epilepsy², the clinical heterogeneity observed in major depressive disorder (MDD) probably stems from the underlying etiological heterogeneity^3,4,5. Recent advances in genome-wide association studies (GWAS) of MDD, facilitated by large samples from the Psychiatric Genomics Consortium^6,7, 23andMe⁸, Million Veteran Program⁹ and global biobanks^10,11,12,13, have yielded substantial progress in identifying MDD-associated genetic variants, and evidence of genetic differences between various clinical subtypes has emerged^3,7,13. Despite these efforts, the quest for subtype-specific genetic loci in MDD has been slow, limiting our understanding of its complex etiology.

In this study, we investigate an important source of MDD heterogeneity, that is, age at onset (AAO), by stratifying individuals into early-onset MDD (eoMDD) and late-onset MDD (loMDD) and conducting a large GWAS meta-analysis of the two subtypes in the ongoing Nordic TRYGGVE collaboration^14,15 (Fig. 1). eoMDD is associated with severe outcomes, including psychotic symptoms, suicidal behavior and comorbidities with other mental disorders and somatic diseases^16,17, whereas loMDD tends to manifest with cognitive decline and increased cardiovascular disease risk¹⁸. Previous attempts to stratify MDD according to AAO have been hindered by methodological challenges, including large variations in AAO across samples, recall bias and relatively small sample sizes¹⁷. To address these challenges, we leveraged the Nordic biobanks and harmonized longitudinal health registries to stratify MDD cases based on age at first MDD diagnosis¹⁵. Previous research suggested that age at first diagnosis could be a useful proxy for AAO, given the high genetic correlation (r_g = 0.95) between the two phenotypes¹⁶.

EMR, electronic medical record.

Full size image

Following harmonization of phenotypic definitions of MDD and age at first diagnosis across nine cohorts from five Nordic countries (Denmark, Estonia, Finland, Norway and Sweden; Methods), we identified 151,582 MDD cases, including 46,708 eoMDD cases with age at first diagnosis of less than 25 years (approximating the 25th percentile of an AAO distribution ≤ 20–21 (ref. ¹⁹; Methods)) and 37,168 loMDD cases with age at first diagnosis of 50 years or older (approximating the 75th percentile of an AAO distribution ≥ 44–45 years; Methods) (Table 1)¹⁹. We conducted harmonized GWAS analyses on eoMDD and loMDD using singularity containers²⁰ in each cohort and then performed meta-analyses. After phenotype harmonization, we observed high genetic correlations (r_g = 0.7–0.9) among the largest Nordic cohorts (the Integrative Psychiatric Research Consortium (iPSYCH), the Estonian Biobank (EstBB), FinnGen) (Supplementary Fig. 1). To assess generalizability outside the Nordic cohorts, we also analyzed UK Biobank (UKB) data, which relies on self-reported age at first diagnosis, and conducted GWAS of eoMDD and loMDD with the same age cutoffs (Supplementary Fig. 1). However, considering the major differences in samples and phenotypes, we conducted primary analyses based on Nordic cohorts of individuals with European ancestry, with the UKB as a comparison cohort for the identified loci.

We identified 12 genome-wide significant loci for eoMDD and two other loci for loMDD (P < 5 × 10⁻⁸) (Fig. 2). These loci were also captured in the GWAS of all cases with MDD, where we identified 80 significant loci (Supplementary Fig. 2 and Supplementary Table 1). Nearly all of these MDD loci have been reported in previous MDD GWAS¹³ (Supplementary Table 3), demonstrating the validity of our harmonized phenotypes and GWAS. For the 17 significant genes identified in the eoMDD analysis, BPTF²¹, PAX5²², SDK1²³ and SORCS3²⁴ are involved in neurodevelopment or synaptic signaling (Supplementary Table 2). For loMDD, we identified four significant genes, with BSN implicated in synaptic neurotransmitter activity²⁵ (Supplementary Table 2). Top loci were only partly replicated in the UKB, with only one locus on chromosome 9 showing nominal significance (P < 0.05); however, for the genome-wide significant loci in either sample, the correlation in their effect sizes was substantial (r = 0.84 for eoMDD; Supplementary Fig. 3 and Supplementary Tables 3–5).

a,b, Mirrored Manhattan plots of the GWAS results from the combined Nordic countries, using inverse-variance-weighted meta-analysis with a genome-wide significance threshold of P < 5 × 10⁻⁸ (denoted by the dashed horizontal line), for eoMDD (a) and loMDD (b). c, Enrichment of open chromatin marks in the eoMDD and loMDD GWAS. The dashed line indicates the one-sided Bonferroni-corrected P threshold, set at P = 0.05/102 = 0.0005. d, Single-nucleotide polymorphism (SNP)-based heritability \(\left({h}_{{\rm{SNP}}}^{2}\right)\) across a range of population prevalence estimates, with the labeled dashed lines indicating the point estimate of the population prevalence. e, Genetic overlap between broad MDD, eoMDD and loMDD, with the upper triangle containing the standard errors and the diagonal containing estimates of polygenicity from SBayesS. * For loMDD, the algorithm did not converge, making the polygenicity estimate unreliable, although it was consistently higher than for eoMDD across successful runs.

Full size image

By integrating GWAS findings on psychiatric disorders with tissue-specific open chromatin marks from the RoadMap Epigenomics Project²⁶, previous research implicated gene regulation during fetal neurodevelopment²⁷. We tested the specific hypothesis that the GWAS of eoMDD would be enriched at regulatory chromatin marks active in fetal brains. Indeed, we found that eoMDD genetic signals were significantly enriched in fetal brain tissues, whereas no enrichment was detected in adult brains (Fig. 2c and Supplementary Table 6a), suggesting a role of early brain development in the risk of eoMDD. For loMDD, we only detected enrichment of one epigenetic marker in male fetal tissues. We did not identify significant enrichment in Genotype-Tissue Expression (GTEx) brain tissues²⁸ or human adult brain cell types²⁹, potentially because of the relatively low statistical power in these subtype-specific GWAS (Supplementary Fig. 4 and Supplementary Table 6b,c).

The single-nucleotide polymorphism (SNP)-based heritability \(\left({h}_{{\rm{SNP}}}^{2}\right)\) estimates for eoMDD were higher than for loMDD at a range of population prevalences (Fig. 2d and Supplementary Table 7). Assuming a population prevalence of 6.2% (Methods), the \({h}_{{\rm{SNP}}}^{2}\) for eoMDD was estimated at 11.2% (95% confidence interval (CI) = 9.9–12.5%; liability scale), which was almost twice as high as that of loMDD at 6% (95% CI = 4.3–6.3%) at a prevalence of 3.2% (Supplementary Table 7). We also conducted a case–case GWAS directly comparing eoMDD with loMDD; however, this comparison showed a small \({h}_{{\rm{SNP}}}^{2}\) (2%, s.e. = 0.74%). Interestingly, analyses of the genetic architecture using SBayesS³⁰ estimated that the polygenicity (that is, the proportion of SNPs with nonzero effects) in eoMDD (4%, posterior s.e. = 0.7%) was much lower than our GWAS of all MDD cases (7%, posterior s.e. = 0.5%) (Fig. 2e) and the estimate reported in the latest Psychiatric Genomics Consortium MDD GWAS (6%)⁶, suggesting that fewer causal variants underpin eoMDD.

The two MDD subtypes correlated moderately (r_g = 0.58, s.e. = 0.04) (Fig. 2e and Supplementary Table 8) and showed differences in their genetic correlations with other traits (Fig. 3a, Supplementary Fig. 5 and Supplementary Table 8). eoMDD had the highest r_g with suicide attempt (r_g = 0.89, s.e. = 0.05), which was more than twice the r_g between loMDD and suicide attempt (r_g = 0.42, s.e. = 0.05) (comparison with r_g in all MDD in Supplementary Fig. 6 and Supplementary Table 8). Similarly, substantial differences were found in their r_g with posttraumatic stress disorder, childhood maltreatment, attention-deficit/hyperactivity disorder, autism spectrum disorder and schizophrenia. For somatic and lifestyle phenotypes, the overlap with eoMDD was significantly stronger for heart failure and body mass index (Supplementary Table 8). Given the r_g between the two MDD subtypes, we further investigated the independent genetic effects of each subtype while conditioning on the other using genomic structural equation modeling (SEM) (Fig. 3b). In this way, the unique overlap between each subtype and the other traits could be evaluated, giving insight into how much of the overlap with the other trait was driven by overlap with the other subtype. After controlling for loMDD, the genetic associations of eoMDD with other traits remained similar, except that the negative r_g between eoMDD and educational attainment was attenuated (Supplementary Table 9). On the other hand, after accounting for the genetic overlap with eoMDD, the genetic associations of loMDD were substantially reduced for many traits (for example, suicide attempt and suicide death), suggesting that the observed r_g between loMDD and these traits were driven by the shared genetics with eoMDD. Overall, eoMDD had stronger genetic overlaps with psychiatric and general health traits than loMDD, with the most notable difference in their genetic correlations with suicide attempt.

a, Genetic correlations from linkage disequilibrium (LD) score regression (LDSC) comparing eoMDD (based on GWAS n = 153,532) and loMDD (GWAS n = 158,588) with key health outcomes and related psychiatric disorders (GWAS n in Supplementary Table 15). The error bars represent the 95% CIs. P values for the associations are reported in Supplementary Table 8. b, Results from genomic SEM linear regression, where the outcomes are regressed on eoMDD (GWAS n = 153,532) while controlling for overlap with loMDD (GWAS n = 158,588), and vice versa. The single asterisks denotes a significant difference between eoMDD and loMDD at P < 0.05. The error bars represent the 95% CIs.

Full size image

We further investigated these relationships using two-sample Mendelian randomization (MR) (Fig. 4). Considering the clear timing difference in the MDD subtypes, we were primarily interested in the role of eoMDD as a risk factor for general health outcomes, while providing the estimates of loMDD as a comparison, and the role of general health traits with loMDD as an outcome. Of note, eoMDD showed a significantly larger effect on suicide attempt than loMDD (eoMDD β = 0.61, s.e. = 0.057; loMDD β = 0.28, s.e. = 0.052), while the magnitude of effects on suicide death was more comparable between the two subtypes (eoMDD β = 0.64, s.e. = 0.20; loMDD β = 0.55, s.e. = 0.21). For educational attainment, our results were consistent with a previous report that suggested lower educational attainment as a putatively causal risk factor for MDD⁷, in particular for loMDD (β = −0.23, s.e. = 0.049; the effect on eoMDD was pleiotropic); furthermore, we showed that eoMDD had a putatively causal effect on lower educational attainment (β = −0.09, s.e. = 0.026). Similarly, our results that both subtypes had a small putatively causal effect on cardiovascular disease (CVD) were in line with previous findings considering MDD as a whole³¹; there was evidence suggesting that heart failure was putatively causal for loMDD. Results from sensitivity analyses were consistent with these main findings (Supplementary Table 10).

a, Results of the MR analyses with MDD as the exposure, with the IVW statistic as the MR effect size estimate (ordered according to the eoMDD effect size). The error bars represent the 95% CIs. b, Effects with MDD as the outcome (ordered according to the effect size in a). Included in both a and b are the number of instrument SNPs used in the analysis (‘SNPs’), the percentage variance explained by the instrument SNPs in the exposure (‘% R²’) and the instrument strength (‘F’). We only tested plausible relationships (in line with time ordering), with improbable relationships, for example, suicide death as a risk factor for MDD, excluded from the figure. If the MR-Egger sensitivity analysis had a significant intercept (triangles), this indicates that the MR estimate was pleiotropic and should not be interpreted. Estimates with an asterisk were significant after correction for multiple testing (exact P values are reported in Supplementary Table 10). The error bars represent the 95% CIs.

Full size image

To investigate whether polygenic risk scores (PRS) for eoMDD and loMDD were associated with clinical indicators or severe outcomes, we generated PRS using leave-one-out (LOO) GWAS summary statistics and tested PRS associations with these outcomes extracted from longitudinal health registry data in each cohort. In the meta-analysis and across individual cohorts, the eoMDD PRS explained a greater proportion of the phenotypic variance compared to the loMDD PRS for all outcomes of interest (Fig. 5a, Supplementary Fig. 7 and Supplementary Table 11). In particular, a 1 s.d. increase in eoMDD PRS was associated with an increased risk of early-onset (odds ratio (OR) = 1.26, 95% CI = 1.13–1.4, P = 1.8 × 10⁻⁵) and lifetime risk of MDD (OR = 1.24, 95% CI = 1.17–1.31, P = 1.16 × 10⁻¹³) compared to the loMDD PRS for early (OR = 1.13, 95% CI = 1.08–1.17, P = 3.9 × 10⁻⁹) and lifetime risk of MDD (OR = 1.13, 95% CI = 1.07–1.19, P = 1.4 × 10⁻⁵). Similarly, compared to the loMDD PRS, the eoMDD PRS was more strongly associated with other outcomes of interest; differences in estimates for eoMDD and loMDD PRS were particularly pronounced for hospitalization and diagnostic conversion to bipolar disorder and schizophrenia.

a, Meta-analyzed associations between PRS and MDD outcomes in the Nordic cohorts, using LOO GWAS summary statistics of eoMDD and loMDD. The error bars represent the 95% CIs. b, Selected mental-health-related and suicide-related ICD-10 group associated with eoMDD and loMDD from the PRS PheWAS. The error bars represent the 95% CIs. c, Number of unique and shared associations with eoMDD and loMDD from the PheWAS. d, Mean symptom scores of self-reported suicidality (Paykel Suicide Scale) stratified according to the top 10%, middle 80% and bottom 10% eoMDD PRS. The error bars represent the 95% CIs. e, Cumulative incidence of suicide attempt over a 10-year period since the first eoMDD diagnosis, stratified according to the top, middle and bottom eoMDD PRS. The gray shading indicates the 95% CIs.

Full size image

To understand how genetic predispositions underlying the two subtypes are associated with comorbidities, we conducted a PRS-based phenome-wide association study (PheWAS) using the EstBB medical records from both primary and specialist care (1,428 International Classification of Diseases, Tenth Revision (ICD-10) code-based diagnosis). After correction for multiple testing (Bonferroni correction, P threshold of 0.05/1,428), we identified both shared and unique associations with either subtype (Fig. 5c and Supplementary Fig. 8a,b). Notably, the eoMDD PRS showed unique associations with psychiatric conditions such as conduct disorder (ICD-10 code: F91) and schizotypal disorder (F21), suicide attempt/intentional self-harm (X78) and problems related to negative life events in childhood (Z61), or other problems related to upbringing (Z62). In contrast, the loMDD PRS was more notably associated with mental and behavioral disorders because of the use of sedatives or hypnotics (F13) and obsessive–compulsive disorder (F42) (Fig. 5b and Supplementary Table 12).

Prompted by our findings of the strong genetic link between eoMDD and suicide attempt, we investigated whether the eoMDD PRS predicts the risk of suicide attempt using self-report surveys and medical records from EstBB. We stratified individuals with eoMDD into three subgroups based on the eoMDD PRS (top and bottom deciles, and middle 80%). First, we observed a dose–response relationship between eoMDD PRS strata and the mean scores of the self-reported Paykel Suicide Scale³², that is, the higher eoMDD PRS, and the higher symptoms of suicide thoughts and attempts (mean symptoms scores = 0.85, 95% CI = 0.79–0.91; 1.05 (1.03–1.07), 1.22 (1.15–1.29) for the bottom, middle and top PRS strata, respectively) (Fig. 5d and Supplementary Table 13a). Next, we estimated the hazard risk ratio (HRR) and the absolute risk for suicide attempt treated in primary or specialist care within a 10-year period since the first recorded MDD diagnosis. Compared with the middle PRS group, individuals in the lowest PRS decile had a significantly lower HRR for suicide attempt of 0.57 (95% CI = 0.49–0.68, P = 1.34 × 10⁻¹⁰), while those in the highest PRS decile had an increased risk (HRR = 1.13, 95% CI = 1.00–1.28, P = 0.058). Furthermore, individuals with the lowest PRS showed a consistently low absolute risk of suicide attempt over time after the first MDD diagnosis (12% cumulative incidence over a 10-year period), while there were no major differences in absolute risk between the middle and highest PRS group until 5.5 years, with the 10-year cumulative incidence at 20% and 26%, respectively (Fig. 5e).

Previous research showed that the incidence of suicide attempts is particularly pronounced during adolescence and young adulthood³³. Early identification of individuals at heightened risk for suicide is of major clinical importance. This period of high incidence of suicide attempts coincides with the risk for eoMDD, making it particularly relevant to perform risk prediction of suicide attempts in this subgroup. Together with our findings suggesting that eoMDD is putatively causal for suicide attempts, these results underscore the potential use of eoMDD PRS in stratifying the risk of suicide attempt among individuals with eoMDD (Supplementary Table 13b) and warrant further investigation of its relevance for suicide prevention³³.

MDD genomics has achieved tremendous progress, from zero significant associations in 2013 to 697 associations in 2025 for the MDD case–control phenotype^6,7,34. In this study, we explored the strategy of going beyond case–control to target specific phenotypic subgroups, aiming to reduce genetic heterogeneity in MDD. Our harmonized healthcare data and analyses across countries improved analytical power and identified differential genetic signals in AAO-based subtypes. A similar approach can be extended to other clinical characteristics, such as vegetative symptoms, psychotic features and disability, to further target clinically relevant subtypes of MDD^3,5,35,36. These findings may have important implications for guiding targeted treatment and prevention in psychiatry.

Study population

To conduct a large GWAS of AAO-based MDD subtypes with consistent phenotypes, we identified nine cohorts across five Nordic countries (Denmark, Sweden, Norway, Finland and Estonia), and a comparison cohort from a non-Nordic country (the UK: UKB). Many of these cohorts are large biobanks linked with lifetime medical records in national patient registers, including iPSYCH, FinnGen, the Norwegian Mother, Father and Child Cohort Study (MoBa), the UKB and the EstBB (cohort details in Supplementary Methods). This study was approved by the Swedish ethical review authority (case no. 2023-03073), and each cohort was approved by the relevant institutional review boards.

Phenotypes

We expanded our previous effort of harmonizing the register-based phenotypes of MDD, age at first diagnosis and its outcomes in the three Scandinavian countries to other Nordic countries with similar registers¹⁵. Using national patient registers, we first extracted information on patient diagnoses of MDD using ICD-10 code F32 (depressive episode) or F33 (recurrent depressive episodes), with the exclusion criteria of a lifetime diagnosis of bipolar disorder or schizophrenia (ICD-10 codes: F30/F31/F32/F33, F20/F23.1/F23.2/F25) (Supplementary Table 14), resulting in a total combined number of MDD cases in the Nordic countries of n = 151,582. Our previous research revealed a high genetic correlation (r_g ≈ 0.95) between AAO and age at first diagnosis for MDD¹⁶, and that onset on average predates first diagnosis by 5 years¹⁶. Therefore, we derived the age at first diagnosis for the MDD patient population and used that as a proxy for AAO for the definition of eoMDD and loMDD. The cutoffs were chosen based on careful review of the literature and the empirical registry data from Sweden. Previous meta-analysis reported that the median AAO for depressive disorder is around age 30, with the 25th and 75th percentile at age 21 and 44, respectively^15,19; our Swedish registry data showed that for age at first MDD specialist diagnosis, the 25th and 75th percentile were at age 25 and 45, respectively¹⁵. Thus, we considered individuals with their first psychiatric specialist treatment contact for MDD at or before age 25 (approximate to the 25th percentile of the AAO distribution at age 20–21) as cases with eoMDD (n = 46,708); cases with loMDD were those with their first specialist treatment contact for MDD at or after age 50 (approximate to the 25th percentile of the AAO distribution at age 44–45; n = 37,168). Controls were individuals without a registered diagnosis of MDD, bipolar disorder or schizophrenia. Because of the differences in cohort design, controls were matched to cases for the population-based cohorts of EstBB, FinnGen, MoBa and UKB. The Danish and Swedish cohorts, which used case–cohort and case–control designs, did not require control matching.

Software containers for federated analyses

To ensure transparent and reproducible analyses of data across sites, we developed software containers and accompanying codes for data processing and analyses (for example, GWAS, PRS, LDSC) and distributed the containers across the study sites. Briefly, the primary software tools and dependencies were installed in virtual machines based on the Ubuntu 20.04 LTS Linux operating system via Docker (https://docker.com) and containerized using the Singularity Image File format (https://sylabs.io) for distribution²⁰. All source codes and files are publicly available via GitHub and released under the GNU General Public License (GPL v.3.0). Tools released under other licenses retain their original licenses.

GWAS and meta-analysis

GWAS of individuals with European ancestries was conducted in each cohort using the REGENIE software package³⁷ available in the software containers²⁰. Analyses were adjusted for the first ten principal components (PCs), age and sex. Summary statistics from each cohort were included in a fixed-effects meta-analysis in METAL³⁸ with the ‘SCHEME STDERR’ command using the inverse variance weight of the corresponding standard errors for all phenotypes studied, resulting in 151,582 cases and 362,873 controls for all MDD, 46,708 cases and 106,824 controls for eoMDD, and 37,168 cases and 121,420 controls for loMDD. Additional variant filtering was applied: INFO score > 0.8, minor allele frequency ≥ 1% and n > 10,000. After this filtering, the numbers of markers for the final METAL analyses were 8,910,578 for broad MDD, 8,848,589 for eoMDD and 8,820,060 for loMDD. Genome-wide significance was set at P < 5 × 10⁻⁸. Genome-wide loci were identified based on genome-wide significant SNPs using PLINK (--clump command). A gene list was then created based on which genes physically overlapped the loci from the previous step (Supplementary Table 2). An additional case–case GWAS was performed comparing eoMDD cases to loMDD cases using the EstBB and FinnGen cohorts, as they include both eoMDD and loMDD cases in their biobanks. A total of 32,060 samples from FinnGen and 23,447 samples from EstBB were meta-analyzed using METAL; heritability \(\left({h}_{{\rm{SNP}}}^{2}\right)\) was estimated using LDSC^39,40.

Gene-based tests

We used MAGMA⁴¹ v.1.08 to test the aggregate association of variants by assigning SNPs to genes. Gene boundaries were expanded by 35 kb upstream and 10 kb downstream to allow for the inclusions of promoters and cis-enhancers. We used the updated deep whole-genome sequencing of the 1000 Genomes Project (European subset) as reference data⁴², downloaded from http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/data_collections/1000G_2504_high_coverage/working/20220422_3202_phased_SNV_INDEL_SV/.

Tissue-type and cell-type enrichment of SNP heritability with stratified LDSC

We used stratified LDSC³⁹ to perform tissue-type and cell-type analysis, estimating the enrichment of SNP heritability in a set of genome annotations. We downloaded precomputed LDSC scores from https://console.cloud.google.com/storage/browser/_details/broad-alkesgroup-public-requester-pays/LDSCORE/LDSC_SEG_ldscores/biorxiv/Roadmap_1000Gv3.tgz, containing data on 396 annotations of epigenetic marks from the ROADMAP project, restricting to the 102 annotations relating to the brain. For the gene expression data, we followed our previous protocol⁴³, first aggregating gene expression across genes to get the average expression per gene in each cell type. We then defined cell-type specificity by dividing the expression of a gene in a cell type by its total expression across all cell types. For each cell type, we considered the top decile of specifically expressed genes as typifying that cell type, and computed LDSC scores for that annotation. In addition to the ROADMAP annotations, we also tested the enrichment of SNP-based heritability (one-sided) in human tissues (37 human tissues from GTEx²⁸) and brain cell types using the latest human single-nucleus transcriptomic data (31 major human brain cell types from ref. ²⁹). We applied a Bonferroni correction for multiple testing in each annotation and trait (for example, in the analysis of ROADMAP data, significance was set at P < 0.05/102 for each trait).

SNP-based heritability and assessing genetic overlap

We used LDSC to estimate SNP-based heritability and genetic correlation⁴⁰. To convert the SNP-based heritability to the liability scale, we calculated the sample prevalence of cases in the meta-analysis; for population prevalence, we used a lifetime estimate of diagnosed MDD of 16% (ref. ⁴⁴). To obtain corresponding estimates for eoMDD and loMDD, we estimated the proportion of cases with MDD who had an early-onset or a late-onset diagnosis in the largest population samples (EstBB and FinnGen) and took that as a percentage of the whole MDD prevalence estimate, resulting in a population estimate of 6.2% for eoMDD and 3.2% for loMDD. Acknowledging that these estimates are approximations, we provided SNP heritability estimates for a range of prevalence estimates (±50% around the point estimates).

To estimate genetic correlations with other traits, we used phenotypes from publicly available summary statistics, selecting important psychiatric disorders and key health determinants and outcomes, including suicide, body mass index, educational attainment, substance use, mortality and cardiovascular disease (GWAS sources in Supplementary Table 15).

SBayesS

Using SBayesS³⁰, we confirmed the LDSC estimates of SNP-based heritability and further estimated the parameter of polygenicity, that is, the proportion of SNPs with nonzero effects, and negative selection. We used the precomputed LD Matrix from UKB participants (‘ukbEURu_imp_v3_HM3_n50k.chisq10’), which is available from the software authors. We ran four chains in parallel to estimate the convergence across starting parameters (--num-chains 4), with each run consisting of 25,000 iterations (--chain-length 25,000), with 5,000 iterations considered as burn-in iterations (--burn-in 5,000).

Genomic SEM

We used genomic SEM to compare the extent to which genetic liability to eoMDD versus loMDD overlaps with genetic liability to other phenotypes⁴⁵. Genomic SEM is an extension of SEM whereby GWAS summary statistics are used to capture the ‘observed’ traits in the model so that the genetic architecture underlying the multivariate relationships between traits can be modeled. Genomic SEM relies on LDSC to estimate genetic correlations and is robust to any amount of sample overlap. We modeled the association of the MDD subtypes with other related phenotypes that have been well studied in the literature, including a range of psychiatric disorders, important health determinants, CVD and mortality (Supplementary Table 15). First, we compared the size and significance of the genetic correlations of the two subtypes (eoMDD and loMDD) with the other phenotypes. Next, we assessed the subtype-specific contribution to the other phenotypes by controlling for the other subtype (for example, we investigated the relationship between eoMDD and suicide attempt after controlling for loMDD). Thus, the unique contribution of each can be gauged while controlling for their shared variance. The genetic overlap between eoMDD and loMDD was moderate; therefore, the concern for strong bias through multicollinearity was limited.

Mendelian randomization

Following the analyses of genetic overlap, we used MR to assess whether the MDD subtypes were causally associated with selected health outcomes (risk factors, mortality and CVD; Supplementary Table 15). We did not include psychiatric outcomes because of the high likelihood of pleiotropy. Assuming that alleles are randomly distributed in the population (in a manner akin to experimental randomization), MR uses genetic variants as instrumental variables to capture a trait (‘exposure’). If the genetic variants are robust, non-pleiotropic instruments for the exposure and there is no direct association with the outcome or an unmeasured confounder, this provides support for a potential causal effect of the exposure. We conducted main and sensitivity analyses as implemented in the TwoSampleMR R package⁴⁶ and computed the IVW statistic to estimate these associations. For interpretation, we focused on effect consistency across tests, rather than P values. Still, to give an indication of the extent of the multiple testing burden, we report significance according to false discovery rate-corrected P values (Benjamini–Hochberg method). We present several sensitivity analyses that correct for potential violations of the MR assumptions. Weighted median and mode analyses are more robust to invalid instruments and correct for outliers⁴⁷. The MR-Egger analysis was used to assess any remaining pleiotropy (intercept) and adjust for such effects. The MR-Egger estimate is often underpowered, but it is still informative to check whether the effect is in the same direction as the other sensitivity analyses⁴⁸. From the same TwoSampleMR package, it is possible to estimate the amount of heterogeneity (Q) in the SNP effects, and estimate the instrument strength F. We added a sensitivity analysis from the MR-PRESSO R package⁴⁹, which has a different and more powerful way to adjust for the effects of outliers and is aimed at correcting for pleiotropy. Finally, we assessed the ‘no measurement error’ assumption by deriving the I² statistic and applied a SIMEX correction on the MR-Egger estimate if I² was between 0.6 and 0.8 (we did not present the MR-Egger estimate when the I² fell below the 0.6 cutoff⁵⁰). Although we were not able to remove potential sample overlap from the analyses (LOO analyses would leave the exposure GWAS underpowered), this should not be a large threat to the validity of our results. The extent of bias stemming from sample overlap is limited if the sample size of the source GWAS is large, the overlap is small and the heritability of the investigated traits is high⁵¹. As shown in Supplementary Table 15, the amount of sample overlap was limited in most cases and present only for a subset of traits.

The aim of the MR analyses was to contrast causal associations between eoMDD and loMDD. However, there was overlap in the instrument sets for both traits. Although only one SNP is present in both sets, many had LD partners in common. Therefore, we performed another sensitivity analysis using a single instrument SNP for each trait. We selected the genome-wide significant SNP that had the largest P value for the other trait (filtering on a minor allele frequency > 0.10). For eoMDD, we used rs7622302 residing in the DAG1 gene on chromosome 3, with P = 0.98 for loMDD. This instrument explained 2.8% of the variance (R²) in eoMDD. For loMDD, we used rs3777421 in the IGF2R gene on chromosome 6, with P = 0.87 for eoMDD. This SNP explained 2.7% variance in loMDD. Although power was expected to be lower in these single-SNP analyses, pleiotropy was also reduced; instrument strength was still sufficient (F = 30.7 for eoMDD and F = 30.1 for loMDD) and the contrast between eoMDD and loMDD was maximized.

Polygenic risk scores

PRS were calculated using LDpred2⁵², which addresses some limitations of LDpred⁵³. Briefly, LDpred2 is a Bayesian method that uses the LD structure in the genome combined with priors regarding the genetic architecture of a trait (SNP heritability and fraction of causal SNPs (hyperparameter p)) to compute PRS. We used the ‘auto’ method of LDpred2 that maximizes the predictive performance of the PRS by testing a range of the fraction of causal SNPs. For our analyses, we used LD matrices provided by the authors of LDpred2⁵², which were generated from the UKB samples (European ancestry) with 1.1 million HapMap3 SNPs. The LDpred2 calculations relied on the bigsnpr R library⁵⁴ and were implemented in a set of custom R scripts⁵⁵.

Individual-level PRS in each cohort were computed using the LOO GWAS summary statistics on MDD subtypes and tested for associations with clinical characteristics or MDD-related outcomes (Supplementary Table 16) in logistic regression models adjusted for at least ten PCs (cohort-dependent), sex and birth year. For the outcomes that were available in all target samples, we performed fixed-effects meta-analysis using the R package metafor⁵⁶.

PRS PheWAS

To investigate the genetic comorbidity of eoMDD and loMDD, we conducted a PRS PheWAS using logistic regression models, using the EstBB electronic health record (EHR) data. Analyses were performed in R (v.4.3). For each ICD-10 code with at least 50 cases (n = 1,151) present in the EstBB EHR, two models were constructed—eoMDD PRS and loMDD PRS as the predictor. These models used the normalized PRS (either eoMDD or loMDD) and a binary outcome variable indicating the presence or absence of the ICD-10 code. The EHR data encompassed both primary and secondary/inpatient care data through linkage to the Estonian central e-health database, covering the time from 2004 to 2022.

The models included sex, year of birth and the first ten genetic PCs as covariates. To account for multiple testing, we applied Bonferroni correction, resulting in a conservative P threshold of 0.05/1,151. ORs (with 95% CIs), the number of cases per ICD-10 code and the P values for both eoMDD and loMDD models are provided in Supplementary Table 12 and Supplementary Figure 8a,b.

Self-reported suicidality stratified according to eoMDD PRS

To evaluate the effect of the eoMDD PRS on self-reported suicidality, we used data from the EstBB Mental Health Online Survey⁵⁷. Self-reported suicidality was assessed using the Paykel Suicide Scale, which includes five items assessing self-reported suicidal thoughts and attempts³². A total of 58,732 individuals who had an MDD diagnosis were present in the EHRs (F32 or F33 codes) and responded to the Paykel Suicide Scale. For this analysis, the first four questions were scored on a scale from 0 to 4 (0 indicating the participant had never felt or experienced this, and 4 indicating the participant felt or experienced it often). The fifth question had binary responses (no/yes) corresponding to values of 0 or 2. The total score of the five items was used for the analysis; the distribution according to eoMDD PRS group is provided in Supplementary Table 13a. The mean values and standard deviations of these sums were then computed for three groups based on eoMDD PRS deciles (1st, 2nd to 9th, 10th). Analyses were conducted using R (v.4.3).

Cox proportional-hazards model of suicide attempt stratified according to eoMDD PRS

To estimate the absolute risk of EHR-based suicide attempt over 10 years after the first diagnosis of MDD, stratified according to eoMDD PRS, we used a Cox proportional-hazards model⁵⁸, using data from the EstBB (results in Supplementary Table 13). Sex and year of birth were included as covariates. The analysis included a cohort of 10,539 individuals diagnosed with MDD (first diagnosis date in 2000 or later, identified by the ICD-10 codes F32 or F33, before or at the age of 25). Participants’ birth years ranged from 1974 to 2004, with a mean birth year of 1990.

We calculated the absolute risk (with 95% CIs) of suicide attempt after the first episode of depression for three groups based on eoMDD polygenic risk score deciles: the 1st decile, the 2nd to 9th deciles and the 10th decile. The risk of suicide attempt was assessed over a 10-year period (3,650 days) after the initial MDD diagnosis. ICD-10 codes X60-X84 and Y87 from the Estonian EHRs were used to identify suicide attempts. The absolute risk was presented as 1 minus the survival probability. All analyses were conducted using R v.4.3, with the ggsurvfit (v.1.0.0)⁵⁸ and survival (v.3.5.7) packages⁵⁹.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

The GWAS summary statistics reported in this article are available via figshare at https://doi.org/10.6084/m9.figshare.27830340 (ref. ⁶⁰).

The central code repository for the software containers for the GWAS and post-GWAS analyses is available via GitHub at https://github.com/comorment/containers and Zenodo at https://doi.org/10.5281/zenodo.15096746 (ref. ⁵⁵); the other tools and public reference data are available via GitHub at https://github.com/comorment/containers.

1. Suzuki, K. et al. Genetic drivers of heterogeneity in type 2 diabetes pathophysiology. Nature 627, 347–357 (2024).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

2. Stevelink, R. et al. GWAS meta-analysis of over 29,000 people with epilepsy identifies 26 risk loci and subtype-specific genetic architecture. Nat. Genet. 55, 1471–1482 (2023).

   Article  Google Scholar 

3. Nguyen, T.-D. et al. Genetic heterogeneity and subtypes of major depression. Mol. Psychiatry 27, 1667–1675 (2022).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

4. Flint, J. The genetic basis of major depressive disorder. Mol. Psychiatry 28, 2254–2265 (2023).

   Article  PubMed  PubMed Central  Google Scholar 

5. Nguyen, T.-D. et al. Genetic contribution to the heterogeneity of major depressive disorder: evidence from a sibling-based design using Swedish national registers. Am. J. Psychiatry 180, 714–722 (2023).

   Article  PubMed  PubMed Central  Google Scholar 

6. Adams, M. J. et al. Trans-ancestry genome-wide study of depression identifies 697 associations implicating cell types and pharmacotherapies. Cell 188, 640–652 (2025).

   Article  Google Scholar 

7. Wray, N. R. et al. Genome-wide association analyses identify 44 risk variants and refine the genetic architecture of major depression. Nat. Genet. 50, 668–681 (2018).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

8. Hyde, C. L. et al. Identification of 15 genetic loci associated with risk of major depression in individuals of European descent. Nat. Genet. 48, 1031–1036 (2016).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

9. Levey, D. F. et al. Bi-ancestral depression GWAS in the Million Veteran Program and meta-analysis in >1.2 million individuals highlight new therapeutic directions. Nat. Neurosci. 24, 954–963 (2021).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

10. Howard, D. M. et al. Genome-wide association study of depression phenotypes in UK Biobank identifies variants in excitatory synaptic pathways. Nat. Commun. 9, 1470 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

11. Yao, X. et al. Integrative analysis of genome-wide association studies identifies novel loci associated with neuropsychiatric disorders. Transl. Psychiatry 11, 69 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

12. Kurki, M. I. et al. FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 613, 508–518 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

13. Als, T. D. et al. Depression pathophysiology, risk prediction of recurrence and comorbid psychiatric disorders using genome-wide analyses. Nat. Med. 29, 1832–1844 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

14. Lu, Y. et al. Genetic risk scores and family history as predictors of schizophrenia in Nordic registers. Psychol. Med. 48, 1201–1208 (2018).

    Article  CAS  PubMed  Google Scholar 

15. Pasman, J. A. et al. Epidemiological overview of major depressive disorder in Scandinavia using nationwide registers. Lancet Reg. Health Eur. 29, 100621 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

16. Harder, A. et al. Genetics of age-at-onset in major depression. Transl. Psychiatry 12, 124 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

17. Power, R. A. et al. Genome-wide association for major depression through age at onset stratification: Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium. Biol. Psychiatry 81, 325–335 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

18. Hashem, A. H., Nasreldin, M., Gomaa, M. A. & Khalaf, O. O. Late versus early onset depression in elderly patients: vascular risk and cognitive impairment. Curr. Aging Sci. 10, 211–216 (2017).

    Article  PubMed  Google Scholar 

19. Solmi, M. et al. Age at onset of mental disorders worldwide: large-scale meta-analysis of 192 epidemiological studies. Mol. Psychiatry 27, 281–295 (2022).

    Article  CAS  PubMed  Google Scholar 

20. Akdeniz, B. C. et al. COSGAP: Containerized statistical genetics analysis pipelines. Bioinform. Adv. 4, vbae067 (2024).

    Article  PubMed  PubMed Central  Google Scholar 

21. Stankiewicz, P. et al. Haploinsufficiency of the chromatin remodeler BPTF causes syndromic developmental and speech delay, postnatal microcephaly, and dysmorphic features. Am. J. Hum. Genet. 101, 503–515 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

22. Gofin, Y. et al. Delineation of a novel neurodevelopmental syndrome associated with PAX5 haploinsufficiency. Hum. Mutat. 43, 461–470 (2022).

    Article  CAS  PubMed  Google Scholar 

23. Yamagata, M. Structure and functions of sidekicks. Front. Mol. Neurosci. 13, 139 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

24. Breiderhoff, T. et al. Sortilin-related receptor SORCS3 is a postsynaptic modulator of synaptic depression and fear extinction. PLoS ONE 8, e75006 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

25. Winter, C. et al. The presynaptic cytomatrix protein Bassoon: sequence and chromosomal localization of the human BSN gene. Genomics 57, 389–397 (1999).

    Article  CAS  PubMed  Google Scholar 

26. Kundaje, A. et al. Integrative analysis of 111 reference human epigenomes. Nature 518, 317–330 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

27. Kim, S. S. et al. Leveraging single-cell ATAC-seq and RNA-seq to identify disease-critical fetal and adult brain cell types. Nat. Commun. 15, 563 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

28. Ardlie, K. G. Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans. Science 348, 648–660 (2015).

    Article  Google Scholar 

29. Siletti, K. et al. Transcriptomic diversity of cell types across the adult human brain. Science 382, eadd7046 (2023).

    Article  CAS  PubMed  Google Scholar 

30. Zeng, J. et al. Widespread signatures of natural selection across human complex traits and functional genomic categories. Nat. Commun. 12, 1164 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

31. Bergstedt, J. et al. Distinct biological signature and modifiable risk factors underlie the comorbidity between major depressive disorder and cardiovascular disease. Nat. Cardiovasc. Res. 3, 754–769 (2024).

    Article  PubMed  PubMed Central  Google Scholar 

32. Paykel, E. S., Myers, J. K., Lindenthal, J. J. & Tanner, J. Suicidal feelings in the general population: a prevalence study. Br. J. Psychiatry 124, 460–469 (1974).

    Article  CAS  PubMed  Google Scholar 

33. Hu, K. et al. Risk factors, impact, and healthcare use related to initial suicide attempts: comprehensive analysis of Swedish population. BMJ Med. 4, e001129 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

34. Ripke, S. et al. A mega-analysis of genome-wide association studies for major depressive disorder. Mol. Psychiatry 18, 497–511 (2013).

    Article  CAS  PubMed  Google Scholar 

35. Nguyen, T.-D. et al. Genetic insights into psychotic major depressive disorder: bridging the mood-psychotic disorder spectrum. EBioMedicine 112, 105576 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

36. Kendler, K. S., Ohlsson, H., Lichtenstein, P., Sundquist, J. & Sundquist, K. The genetic epidemiology of treated major depression in Sweden. Am. J. Psychiatry 175, 1137–1144 (2018).

    Article  PubMed  Google Scholar 

37. Mbatchou, J. et al. Computationally efficient whole-genome regression for quantitative and binary traits. Nat. Genet. 53, 1097–1103 (2021).

    Article  CAS  PubMed  Google Scholar 

38. Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26, 2190–2191 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

39. Finucane, H. K. et al. Partitioning heritability by functional annotation using genome-wide association summary statistics. Nat. Genet. 47, 1228–1235 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

40. Bulik-Sullivan, B. K. et al. LD Score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291–295 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

41. de Leeuw, C. A., Mooij, J. M., Heskes, T. & Posthuma, D. MAGMA: generalized gene-set analysis of GWAS data. PLoS Comput. Biol. 11, e1004219 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

42. Byrska-Bishop, M. et al. High-coverage whole-genome sequencing of the expanded 1000 Genomes Project cohort including 602 trios. Cell 185, 3426–3440 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

43. Skene, N. G. et al. Genetic identification of brain cell types underlying schizophrenia. Nat. Genet. 50, 825–833 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

44. Kessler, R. C., Chiu, W. T., Demler, O., Merikangas, K. R. & Walters, E. E. Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey Replication. Arch. Gen. Psychiatry 62, 617–627 (2005).

    Article  PubMed  PubMed Central  Google Scholar 

45. Grotzinger, A. D. et al. Genomic structural equation modelling provides insights into the multivariate genetic architecture of complex traits. Nat. Hum. Behav. 3, 513–525 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

46. Hemani, G. et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife 7, e34408 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

47. Hartwig, F. P., Davey Smith, G. & Bowden, J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int. J. Epidemiol. 46, 1985–1998 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

48. Bowden, J., Davey Smith, G. & Burgess, S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int. J. Epidemiol. 44, 512–525 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

49. Verbanck, M., Chen, C.-Y., Neale, B. & Do, R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat. Genet. 50, 693–698 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

50. Bowden, J. et al. Assessing the suitability of summary data for two-sample Mendelian randomization analyses using MR-Egger regression: the role of the I2 statistic. Int. J. Epidemiol. 45, 1961–1974 (2016).

    PubMed  PubMed Central  Google Scholar 

51. Mounier, N. & Kutalik, Z. Bias correction for inverse variance weighting Mendelian randomization. Genet. Epidemiol. 47, 314–331 (2023).

    Article  CAS  PubMed  Google Scholar 

52. Privé, F., Arbel, J. & Vilhjálmsson, B. J. LDpred2: better, faster, stronger. Bioinformatics 36, 5424–5431 (2020).

    Article  PubMed Central  Google Scholar 

53. Vilhjálmsson, B. J. et al. Modeling linkage disequilibrium increases accuracy of polygenic risk scores. Am. J. Hum. Genet. 97, 576–592 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

54. Privé, F., Aschard, H., Ziyatdinov, A. & Blum, M. G. B. Efficient analysis of large-scale genome-wide data with two R packages: bigstatsr and bigsnpr. Bioinformatics 34, 2781–2787 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

55. Frei, O. et al. comorment/containers: Comorment-Containers-v1.9.1. Zenodo https://doi.org/10.5281/zenodo.15096746 (2025).

56. Viechtbauer, W. Conducting meta-analyses in R with the metafor package. J. Stat. Softw. 36, 1–48 (2010).

    Article  Google Scholar 

57. Ojalo, T. et al. Cohort profile update: Mental Health Online Survey in the Estonian Biobank (EstBB MHoS). Int. J. Epidemiol. 53, dyae017 (2024).

    Article  PubMed  PubMed Central  Google Scholar 

58. Sjöberg, D. D., Baillie, M., Fruechtenicht, C., Haesendonckx, S. & Treis, T. ggsurvfit: Flexible time-to-event figures. R version 4.3 www.danieldsjoberg.com/ggsurvfit/ (2024).

59. Therneau, T. M., Atkinson, E. & Crowson, C. survival: Survival Analysis. R version 4.3 https://cran.r-project.org/web/packages/survival/survival.pdf (2025).

60. Lu, Yi. GWAS sumstats of early-onset, late-onset, and all MDD. figshare https://doi.org/10.6084/m9.figshare.27830340 (2025).

Download references

Author notes

1. These authors contributed equally: John R. Shorter, Joëlle A. Pasman, Siim Kurvits, Andreas Jangmo, Joonas Naamanka.

2. These authors jointly supervised this work: Martin Tesli, Kelli Lehto, Alfonso Buil, Yi Lu.

3. A full list of members and their affiliations appears in the Supplementary Information.

Authors and Affiliations

1. Department of Science and Environment, Roskilde University, Roskilde, Denmark

   John R. Shorter

2. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden

   Joëlle A. Pasman, Arvid Harder, Kaarina Kowalec, Ying Xiong, Sara Hägg, Mikael Landén, Patrick F. Sullivan & Yi Lu

3. Department of Psychiatry, Amsterdam UMC location University of Amsterdam, Amsterdam, the Netherlands

   Joëlle A. Pasman

4. Estonian Genome Centre, Institute of Genomics, University of Tartu, Tartu, Estonia

   Siim Kurvits & Kelli Lehto

5. Department of Chronic Diseases, Norwegian Institute of Public Health, Oslo, Norway

   Andreas Jangmo

6. SleepWell Research Program and Department of Psychology, Faculty of Medicine, University of Helsinki, Helsinki, Finland

   Joonas Naamanka, Nelli Frilander & Iiris Hovatta

7. Central Institute of Mental Health, Hector Institute for Artificial Intelligence in Psychiatry, Mannheim, Germany

   Joonas Naamanka

8. Centre for Precision Psychiatry, Institute of Clinical Medicine, University of Oslo and Oslo University Hospital, Oslo, Norway

   Espen Hagen, Elise Koch, Bayram C. Akdeniz, Oleksandr Frei & Ole A. Andreassen

9. College of Pharmacy, University of Manitoba, Winnipeg, Manitoba, Canada

   Kaarina Kowalec

10. Department of Biochemistry and Medical Genetics, University of Manitoba, Winnipeg, Manitoba, Canada

    Kaarina Kowalec

11. Institute of Biological Psychiatry, Mental Health Center Sct. Hans, Mental Health Services Copenhagen, Copenhagen University Hospital, Roskilde, Denmark

    Richard Zetterberg, Joeri J. Meijsen, Jesper R. Gådin, Thomas M. Werge & Alfonso Buil

12. Center for Eating and feeding Disorders Research, Mental Health Center Ballerup, Copenhagen University Hospital, Mental Health Services Copenhagen, Copenhagen, Denmark

    Joeri J. Meijsen

13. Unit of Integrative Epidemiology, Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden

    Jacob Bergstedt

14. Institute of Neuroscience and Physiology, University of Gothenburg, Gothenburg, Sweden

    Mikael Landén

15. Centre for Psychiatry Research, Department of Clinical Neuroscience, Karolinska Institutet & Stockholm Health Care Services, Region Stockholm, Karolinska Institutet, Stockholm, Sweden

    Christian Rück & John Wallert

16. Department of Women’s and Children’s Health, Uppsala University, Uppsala, Sweden

    Alkistis Skalkidou

17. Department of Mental Health and Suicide, Norwegian Institute of Public Health, Oslo, Norway

    Ted Reichborn-Kjennerud

18. Institute of Clinical Medicine, University of Oslo, Oslo, Norway

    Ted Reichborn-Kjennerud & Martin Tesli

19. Department of Clinical Medicine, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark

    Thomas M. Werge

20. Departments of Genetics and Psychiatry, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA

    Patrick F. Sullivan

21. KG Jebsen Centre for Neurodevelopmental Disorders, University of Oslo, Oslo, Norway

    Ole A. Andreassen

22. Division of Mental Health and Substance Abuse, Diakonhjemmet Hospital, Oslo, Norway

    Martin Tesli

23. Department of Mental Health, Norwegian Institute of Public Health, Oslo, Norway

    Martin Tesli

24. Department of Psychiatry, Østfold Hospital, Grålum, Norway

    Martin Tesli

Consortia

Contributions

J.R.S. and Y.L. coordinated the project. The core analytical team consisted of J.R.S., J.A.P., S.K., A.J. and J.N. M.T., K.L., A.B. and Y.L. proposed and designed the study. The core writing group consisted of J.R.S., J.A.P., S.K., A.J., J.N., M.T., K.L., A.B. and Y.L. J.R.S., J.A.P., S.K., A.J., J.N., A.H., E.H., K.K., N.F., R.Z., J.J.M., J.R.G., J.B., Y.X., E.K., B.C.A. and O.F. contributed to the data analysis or interpretation of the results. S.H., M.L., C.R., J.W., A.S., I.H., T.R.-K., T.M.W., P.F.S., O.A.A., M.T., K.L., A.B. and Y.L. contributed to the data or obtained funding. T.M.W., P.F.S., O.A.A., M.T., K.L., A.B. and Y.L. helped to conceive the study and supervised the project. All authors commented on the manuscript and approved the final version.

Corresponding authors

Correspondence to John R. Shorter or Yi Lu.

Competing interests

P.F.S. was a paid adviser and shareholder for Neumora Therapeutics. O.A.A. has received speaker fees from Lundbeck, Janssen, Otsuka and Sunovion, and is a consultant to Cortechs.ai and Precision Health. I.H. has received speaker honoraria from Lundbeck and Otsuka. The other authors declare no competing interests.

Peer review information

Nature Genetics thanks Nick Bass, Na Cai and Christel Middeldorp for their contribution to the peer review of this work. Peer reviewer reports are available.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions