Genomic SEM applied to explore etiological divergences in bipolar subtypes.

BACKGROUND: Bipolar disorder (BD) is an overarching diagnostic class defined by the presence of at least one prior manic episode (BD I) or both a prior hypomanic episode and a prior depressive episode (BD II). Traditionally, BD II has been conceptualized as a less severe presentation of BD I, however, extant literature to investigate this claim has been mixed. METHODS: We apply genomic structural equation modeling (Genomic SEM) to investigate divergent genetic pathways across BD's two major subtypes using the most recent GWAS summary statistics from the PGC. We begin by identifying divergences in genetic correlations across 98 external traits using a Bonferroni-corrected threshold. We also use a theoretically informed follow-up model to examine the extent to which the genetic variance in each subtype is explained by schizophrenia and major depression. Lastly, transcriptome-wide SEM (T-SEM) was used to identify neuronal gene expression patterns associated with BD subtypes. RESULTS: BD II was characterized by significantly larger genetic overlap across non-psychiatric medical and internalizing traits (e.g. heart disease, neuroticism, insomnia), while stronger associations for BD I were absent. Consistent with these findings, follow-up modeling revealed a substantial major depression component for BD II. T-SEM results revealed 35 unique genes associated with shared risk across BD subtypes. CONCLUSIONS: Divergent patterns of genetic relationships across external traits provide support for the distinction of the bipolar subtypes. However, our results also challenge the illness severity conceptualization of BD given stronger genetic overlap across BD II and a range of clinically relevant traits and disorders.

Genome-wide meta-analysis of ascertainment and symptom structures of major depression in case-enriched and community cohorts.

BACKGROUND: Diagnostic criteria for major depressive disorder allow for heterogeneous symptom profiles but genetic analysis of major depressive symptoms has the potential to identify clinical and etiological subtypes. There are several challenges to integrating symptom data from genetically informative cohorts, such as sample size differences between clinical and community cohorts and various patterns of missing data. METHODS: We conducted genome-wide association studies of major depressive symptoms in three cohorts that were enriched for participants with a diagnosis of depression (Psychiatric Genomics Consortium, Australian Genetics of Depression Study, Generation Scotland) and three community cohorts who were not recruited on the basis of diagnosis (Avon Longitudinal Study of Parents and Children, Estonian Biobank, and UK Biobank). We fit a series of confirmatory factor models with factors that accounted for how symptom data was sampled and then compared alternative models with different symptom factors. RESULTS: The best fitting model had a distinct factor for Appetite/Weight symptoms and an additional measurement factor that accounted for the skip-structure in community cohorts (use of Depression and Anhedonia as gating symptoms). CONCLUSION: The results show the importance of assessing the directionality of symptoms (such as hypersomnia versus insomnia) and of accounting for study and measurement design when meta-analyzing genetic association data.

Partitioning the Genomic Components of Behavioral Disinhibition and Substance Use (Disorder) Using Genomic Structural Equation Modeling.

Externalizing behaviors encompass manifestations of risk-taking, self-regulation, aggression, sensation-/reward-seeking, and impulsivity. Externalizing research often includes substance use (SUB), substance use disorder (SUD), and other (non-SUB/SUD) "behavioral disinhibition" (BD) traits. Genome-wide and twin research have pointed to overlapping genetic architecture within and across SUB, SUD, and BD. We created single-factor measurement models-each describing SUB, SUD, or BD traits-based on mutually exclusive sets of European ancestry genome-wide association study (GWAS) statistics exploring externalizing variables. We then assessed the partitioning of genetic covariance among the three facets using correlated factors models and Cholesky decomposition. Even when the residuals for indicators relating to the same substance were correlated across the SUB and SUD factors, the two factors yielded a large correlation (rg = 0.803). BD correlated strongly with the SUD (rg = 0.774) and SUB (rg = 0.778) factors. In our initial decompositions, 33% of total BD variance remained after partialing out SUD and SUB. The majority of covariance between BD and SUB and between BD and SUD was shared across all factors, and, within these models, only a small fraction of the total variation in BD operated via an independent pathway with SUD or SUB outside of the other factor. When only nicotine/tobacco, cannabis, and alcohol were included for the SUB/SUD factors, their correlation increased to rg = 0.861; in corresponding decompositions, BD-specific variance decreased to 27%. Further research can better elucidate the properties of BD-specific variation by exploring its genetic/molecular correlates.

Disentangling differing relationships between internalizing disorders and alcohol use.

Both internalizing disorders and alcohol use have dramatic, wide-spread implications for global health. Previous work has established common phenotypic comorbidity among these disorders, as well as shared genetic variation underlying them both. We used genomic structural equation modeling to investigate the shared genetics of internalizing, externalizing, and alcohol use traits, as well as to explore whether specific domains of internalizing symptoms mediate the contrasting relationships with problematic alcohol use compared to alcohol consumption. We also examined patterns of genetic correlations between similar traits within additional Finnish and East Asian ancestry groups. When the shared genetic influence of externalizing psychopathology was accounted for, the genetic effect of internalizing traits on alcohol use was reduced, suggesting the important role of common genetic factors underlying multiple psychiatric disorders and their genetic influences on comorbidity of internalizing and alcohol use traits. Individual internalizing domains had contrasting effects on frequency of alcohol consumption, which demonstrate the complex system of pleiotropy that exists, even within similar disorders, and can be missed when evaluating only relationships among formal diagnoses. Future work must consider the broad effects of shared psychopathology along with the fine-scale effects of heterogeneity within disorders to more fully understand the biology underlying complex traits.

Using Genomic Structural Equation Modeling to Partition the Genetic Covariance Between Birthweight and Cardiometabolic Risk Factors into Maternal and Offspring Components in the Norwegian HUNT Study.

The Barker Hypothesis posits that adverse intrauterine environments result in fetal growth restriction and increased risk of cardiometabolic disease through developmental compensations. Here we introduce a new statistical model using the genomic SEM software that is capable of simultaneously partitioning the genetic covariation between birthweight and cardiometabolic traits into maternally mediated and offspring mediated contributions. We model the covariance between birthweight and later life outcomes, such as blood pressure, non-fasting glucose, blood lipids and body mass index in the Norwegian HUNT study, consisting of 15,261 mother-eldest offspring pairs with genetic and phenotypic data. Application of this model showed some evidence for maternally mediated effects of systolic blood pressure on offspring birthweight, and pleiotropy between birthweight and non-fasting glucose mediated through the offspring genome. This underscores the importance of genetic links between birthweight and cardiometabolic phenotypes and offer alternative explanations to environmentally based hypotheses for the phenotypic correlation between these variables.

Guidelines for Evaluating the Comparability of Down-Sampled GWAS Summary Statistics.

Proprietary genetic datasets are valuable for boosting the statistical power of genome-wide association studies (GWASs), but their use can restrict investigators from publicly sharing the resulting summary statistics. Although researchers can resort to sharing down-sampled versions that exclude restricted data, down-sampling reduces power and might change the genetic etiology of the phenotype being studied. These problems are further complicated when using multivariate GWAS methods, such as genomic structural equation modeling (Genomic SEM), that model genetic correlations across multiple traits. Here, we propose a systematic approach to assess the comparability of GWAS summary statistics that include versus exclude restricted data. Illustrating this approach with a multivariate GWAS of an externalizing factor, we assessed the impact of down-sampling on (1) the strength of the genetic signal in univariate GWASs, (2) the factor loadings and model fit in multivariate Genomic SEM, (3) the strength of the genetic signal at the factor level, (4) insights from gene-property analyses, (5) the pattern of genetic correlations with other traits, and (6) polygenic score analyses in independent samples. For the externalizing GWAS, although down-sampling resulted in a loss of genetic signal and fewer genome-wide significant loci; the factor loadings and model fit, gene-property analyses, genetic correlations, and polygenic score analyses were found robust. Given the importance of data sharing for the advancement of open science, we recommend that investigators who generate and share down-sampled summary statistics report these analyses as accompanying documentation to support other researchers' use of the summary statistics.

Multivariate estimation of factor structures of complex traits using SNP-based genomic relationships.

BACKGROUND: Heritability and genetic correlation can be estimated from genome-wide single-nucleotide polymorphism (SNP) data using various methods. We recently developed multivariate genomic-relatedness-based restricted maximum likelihood (MGREML) for statistically and computationally efficient estimation of SNP-based heritability ([Formula: see text]) and genetic correlation ([Formula: see text]) across many traits in large datasets. Here, we extend MGREML by allowing it to fit and perform tests on user-specified factor models, while preserving the low computational complexity. RESULTS: Using simulations, we show that MGREML yields consistent estimates and valid inferences for such factor models at low computational cost (e.g., for data on 50 traits and 20,000 individuals, a saturated model involving 50 [Formula: see text]'s, 1225 [Formula: see text]'s, and 50 fixed effects is estimated and compared to a restricted model in less than one hour on a single notebook with two 2.7 GHz cores and 16 GB of RAM). Using repeated measures of height and body mass index from the US Health and Retirement Study, we illustrate the ability of MGREML to estimate a factor model and test whether it fits the data better than a nested model. The MGREML tool, the simulation code, and an extensive tutorial are freely available at https://github.com/devlaming/mgreml/ . CONCLUSION: MGREML can now be used to estimate multivariate factor structures and perform inferences on such factor models at low computational cost. This new feature enables simple structural equation modeling using MGREML, allowing researchers to specify, estimate, and compare genetic factor models of their choosing using SNP data.

Shared genetic influences on resting-state functional networks of the brain.

The amplitude of activation in brain resting state networks (RSNs), measured with resting-state functional magnetic resonance imaging, is heritable and genetically correlated across RSNs, indicating pleiotropy. Recent univariate genome-wide association studies (GWASs) explored the genetic underpinnings of individual variation in RSN activity. Yet univariate genomic analyses do not describe the pleiotropic nature of RSNs. In this study, we used a novel multivariate method called genomic structural equation modeling to model latent factors that capture the shared genomic influence on RSNs and to identify single nucleotide polymorphisms (SNPs) and genes driving this pleiotropy. Using summary statistics from GWAS of 21 RSNs reported in UK Biobank (N = 31,688), the genomic latent factor analysis was first conducted in a discovery sample (N = 21,081), and then tested in an independent sample from the same cohort (N = 10,607). In the discovery sample, we show that the genetic organization of RSNs can be best explained by two distinct but correlated genetic factors that divide multimodal association networks and sensory networks. Eleven of the 17 factor loadings were replicated in the independent sample. With the multivariate GWAS, we found and replicated nine independent SNPs associated with the joint architecture of RSNs. Further, by combining the discovery and replication samples, we discovered additional SNP and gene associations with the two factors of RSN amplitude. We conclude that modeling the genetic effects on brain function in a multivariate way is a powerful approach to learn more about the biological mechanisms involved in brain function.

Examining the Genetic Links between Clusters of Immune-mediated Diseases and Psychiatric Disorders.

Importance: Autoimmune and autoinflammatory diseases have been linked to psychiatric disorders in the phenotypic and genetic literature. However, a comprehensive model that investigates the association between a broad range of psychiatric disorders and immune-mediated disease in a multivariate framework is lacking. Objective: This study aims to establish a factor structure based on the genetic correlations of immune-mediated diseases and investigate their genetic relationships with clusters of psychiatric disorders. Design Setting and Participants: We utilized Genomic Structural Equation Modeling (Genomic SEM) to establish a factor structure of 11 immune-mediated diseases. Genetic correlations between these immune factors were examined with five established factors across 13 psychiatric disorders representing compulsive, schizophrenia/bipolar, neurodevelopmental, internalizing, and substance use disorders. We included GWAS summary statistics of individuals of European ancestry with sample sizes from 1,223 cases for Addison's disease to 170,756 cases for major depressive disorder. Main Outcomes and Measures: Genetic correlations between psychiatric and immune-mediated disease factors and traits to determine genetic overlap. We develop and validate a new heterogeneity metric, Q Factor , that quantifies the degree to which factor correlations are driven by more specific pairwise associations. We also estimate residual genetic correlations between pairs of psychiatric disorders and immune-mediated diseases. Results: A four-factor model of immune-mediated diseases fit the data well and described a continuum from autoimmune to autoinflammatory diseases. The four factors reflected autoimmune, celiac, mixed pattern, and autoinflammatory diseases. Analyses revealed seven significant factor correlations between the immune and psychiatric factors, including autoimmune and mixed pattern diseases with the internalizing and substance use factors, and autoinflammatory diseases with the compulsive, schizophrenia/bipolar, and internalizing factors. Additionally, we find evidence of divergence in associations within factors as indicated by Q Factor . This is further supported by 14 significant residual genetic correlations between individual psychiatric disorders and immune-mediated diseases. Conclusion and Relevance: Our results revealed genetic links between clusters of immune-mediated diseases and psychiatric disorders. Current analyses indicate that previously described relationships between specific psychiatric disorders and immune-mediated diseases often capture broader pathways of risk sharing indexed by our genomic factors, yet are more specific than a general association across all psychiatric disorders and immune-mediated diseases.

Examining Differences in the Genetic and Functional Architecture of Attention-Deficit/Hyperactivity Disorder Diagnosed in Childhood and Adulthood.

Background: Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental disorder with diagnostic criteria requiring symptoms to begin in childhood. We investigated whether individuals diagnosed as children differ from those diagnosed in adulthood with respect to shared and unique architecture at the genome-wide and gene expression level of analysis. Methods: We used genomic structural equation modeling (SEM) to investigate differences in genetic correlations (rg) of childhood-diagnosed (ncases = 14,878) and adulthood-diagnosed (ncases = 6961) ADHD with 98 behavioral, psychiatric, cognitive, and health outcomes. We went on to apply transcriptome-wide SEM to identify functional annotations and patterns of gene expression associated with genetic risk sharing or divergence across the ADHD subgroups. Results: Compared with the childhood subgroup, adulthood-diagnosed ADHD exhibited a significantly larger negative rg with educational attainment, the noncognitive skills of educational attainment, and age at first sexual intercourse. We observed a larger positive rg for adulthood-diagnosed ADHD with major depression, suicidal ideation, and a latent internalizing factor. At the gene expression level, transcriptome-wide SEM analyses revealed 22 genes that were significantly associated with shared genetic risk across the subtypes that reflected a mixture of coding and noncoding genes and included 15 novel genes relative to the ADHD subgroups. Conclusions: This study demonstrated that ADHD diagnosed later in life shows much stronger genetic overlap with internalizing disorders and related traits. This may indicate the potential clinical relevance of distinguishing these subgroups or increased misdiagnosis for those diagnosed later in life. Top transcriptome-wide SEM results implicated genes related to neuronal function and clinical characteristics (e.g., sleep).

It is unclear whether individuals who are diagnosed with attention-deficit/hyperactivity disorder (ADHD) as children differ from those who are diagnosed in adulthood with respect to their genetic architecture. We found that adulthood-diagnosed ADHD is much more genetically similar than ADHD diagnosed in childhood to disorders in the internalizing space, such as depression and suicidality. Differences between the distinct age groups at diagnosis highlight the importance of distinguishing these subgroups in a clinical and treatment setting.

Executive Function and Impulsivity Predict Distinct Genetic Variance in Internalizing Problems, Externalizing Problems, Thought Disorders, and Compulsive Disorders: A Genomic Structural Equation Modeling Study.

Individual differences in self-control predict many health and life outcomes. Building on twin literature, we used genomic structural equation modeling to test the hypothesis that genetic influences on executive function and impulsivity predict independent variance in mental health and other outcomes. The impulsivity factor (comprising urgency, lack of premeditation, and other facets) was only modestly genetically correlated with low executive function (rg =.13). Controlling for impulsivity, low executive function was genetically associated with increased internalizing (ßg =.15), externalizing (ßg =.13), thought disorders (ßg =.38), compulsive disorders (ßg =.22), and chronotype (ßg =.11). Controlling for executive function, impulsivity was positively genetically associated with internalizing (ßg =.36), externalizing (ßg =.55), body mass index (ßg =.26), and insomnia (ßg =.35), and negatively genetically associated with compulsive disorders (ßg = -.17). Executive function and impulsivity were both genetically correlated with general cognitive ability and educational attainment. This work suggests that executive function and impulsivity are genetically separable and show independent associations with mental health.

Influences on the Genetic Relationship Between Cannabis Use and Schizophrenia: The Role of the Externalizing Spectrum.

BACKGROUND AND HYPOTHESIS: The nature of the robust association between cannabis use and schizophrenia remains undetermined. Plausible hypotheses explaining this relationship include the premise that cannabis use causes schizophrenia, increased liability for schizophrenia increases the risk of cannabis use initiation (eg, self-medication), or the bidirectional causal hypothesis where both factors play a role in the development of the other. Alternatively, factors that confound the relationship between schizophrenia and cannabis use may explain their association. Externalizing behaviors are related to both schizophrenia and cannabis use and may influence their relationship. STUDY DESIGN: This study aimed to evaluate whether externalizing behaviors influence the genetic relationship between cannabis use and schizophrenia. We conducted a multivariate genome-wide association analysis of 6 externalizing behaviors in order to construct a genetic latent factor of the externalizing spectrum. Genomic structural equation modeling was used to evaluate the influence of externalizing behaviors on the genetic relationship between cannabis use and schizophrenia. RESULTS: We found that externalizing behaviors partially explained the association between cannabis use and schizophrenia by up to 42%. CONCLUSIONS: This partial explanation of the association by externalizing behaviors suggests that there may be other unidentified confounding factors, alongside a possible direct association between schizophrenia and cannabis use. Future studies should aim to identify further confounding factors to accurately explain the relationship between cannabis use and schizophrenia.

The shared genetic architecture of modifiable risk for Alzheimer's disease: a genomic structural equation modelling study.

Targeting modifiable risk factors may help to prevent Alzheimer's disease (AD), but the pathways by which these risk factors influence AD risk remain incompletely understood. We identified genome-wide association studies for AD and its major modifiable risk factors. We calculated the genetic correlation among these traits and modelled this using genomic structural equation modelling. We identified complex networks of genetic overlap among AD risk factors, but AD itself was largely genetically distinct. The data were best explained by a bi-factor model, incorporating a Common Factor for AD risk, and 3 orthogonal sub-clusters of risk factors. Taken together, our findings suggest that there is extensive shared genetic architecture between AD modifiable risk factors, but this is largely independent of AD genetic pathways. Extensive genetic pleiotropy between risk factors may influence AD indirectly by decreasing cognitive reserve or increasing risk of multimorbidity, leading to poorer brain health. Further work to understand the biology reflected by this communality may provide novel mechanistic insights that could help to prioritise targets for dementia prevention.

Cloud computing for voxel-wise SEM analysis of MRI data.

As data collection costs fall and vast quantities of data are collected, data analysis time can become a bottleneck. For massively parallel analyses, cloud computing offers the short-term rental of ample processing power. Recent software innovations have reduced the offort needed to take advantage of cloud computing. To demonstrate, we replicate a voxel-wise examination of the genetic contributions to cortical development by age using evidence from 1,748 MRI scans. Specifically, we employ off-the-shelf Kubernetes software that permits us to re-run our analyses using almost the same computer code as was published in the original article. Large, well funded institutions may continue to maintain their own computing clusters. However, the modest cost of renting and ease of utilizing cloud computing services makes unprecedented compute power available to all researchers, whether or not affliated with a research institution. We expect this to spur innovation in the sophisticated modeling of large datasets.