Dataset Information

The predictive capacity of psychiatric and psychological polygenic risk scores for distinguishing cases in a child and adolescent psychiatric sample from controls.

ABSTRACT:

Background

Psychiatric traits are heritable, highly comorbid and genetically correlated, suggesting that genetic effects that are shared across disorders are at play. The aim of the present study is to quantify the predictive capacity of common genetic variation of a variety of traits, as captured by their PRS, to predict case-control status in a child and adolescent psychiatric sample including controls to reveal which traits contribute to the shared genetic risk across disorders.

Method

Polygenic risk scores (PRS) of 14 traits were used as predictor phenotypes to predict case-control status in a clinical sample. Clinical cases (N = 1,402), age 1-21, diagnostic categories: Autism spectrum disorders (N = 492), Attention-deficit/ hyperactivity disorders (N = 471), Anxiety (N = 293), disruptive behaviors (N = 101), eating disorders (N = 97), OCD (N = 43), Tic disorder (N = 50), Disorder of infancy, childhood or adolescence NOS (N = 65), depression (N = 64), motor, learning and communication disorders (N = 59), Anorexia Nervosa (N = 48), somatoform disorders (N = 47), Trauma/stress (N = 39) and controls (N = 1,448, age 17-84) of European ancestry. First, these 14 PRS were tested in univariate regression analyses. The traits that significantly predicted case-control status were included in a multivariable regression model to investigate the gain in explained variance when leveraging the genetic effects of multiple traits simultaneously.

Results

In the univariate analyses, we observed significant associations between clinical status and the PRS of educational attainment (EA), smoking initiation (SI), intelligence, neuroticism, alcohol dependence, ADHD, major depression and anti-social behavior. EA (p-value: 3.53E-20, explained variance: 3.99%, OR: 0.66), and SI (p-value: 4.77E-10, explained variance: 1.91%, OR: 1.33) were the most predictive traits. In the multivariable analysis with these eight significant traits, EA and SI, remained significant predictors. The explained variance of the PRS in the model with these eight traits combined was 5.9%.

Conclusion

Our study provides more insights into the genetic signal that is shared between childhood and adolescent psychiatric disorders. As such, our findings might guide future studies on psychiatric comorbidity and offer insights into shared etiology between psychiatric disorders. The increase in explained variance when leveraging the genetic signal of different predictor traits supports a multivariable approach to optimize precision accuracy for general psychopathology.

SUBMITTER: Jansen AG

PROVIDER: S-EPMC8453516 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Sleep disturbance is a common symptom of psychiatric and neurodevelopmental disorders and, especially in childhood, can be a precursor to various mental disorders. However, the genetic etiology of mental illness that contributes to sleep disturbance during childhood is poorly understood. We investigated whether the polygenic features of psychiatric and neurodevelopmental disorders are associated with sleep disturbance during childhood. We conducted polygenic risk score (PRS) analyses by utilizing large-scale genome-wide association studies (GWASs) (n = 46,350-500,199) of five major psychiatric and neurodevelopmental disorders (autism spectrum disorder, schizophrenia, attention-deficit/hyperactivity disorder (ADHD), major depressive disorder (MDD), and bipolar disorder) and, additionally, anxiety disorders as base datasets. We used the data of 9- to 10-year-olds from the Adolescent Brain Cognitive Development study (n = 9683) as a target dataset. Sleep disturbance was assessed based on the Sleep Disturbance Scale for Children (SDSC) scores. The effects of PRSs for these psychiatric and neurodevelopmental disorders on the total scores and six subscale scores of the SDSC were investigated. Of the PRSs for the five psychiatric and neurodevelopmental disorders, the PRSs for ADHD and MDD positively correlated with sleep disturbance in children (ADHD: R2 = 0.0033, p = 6.19 × 10-5, MDD: R2 = 0.0042, p = 5.69 × 10-6). Regarding the six subscale scores of the SDSC, the PRSs for ADHD positively correlated with both disorders of initiating and maintaining sleep (R2 = 0.0028, p = 2.31 × 10-4) and excessive somnolence (R2 = 0.0023, p = 8.44 × 10-4). Furthermore, the PRSs for MDD primarily positively correlated with disorders of initiating and maintaining sleep (R2 = 0.0048, p = 1.26 × 10-6), followed by excessive somnolence (R2 = 0.0023, p = 7.74 × 10-4) and sleep hyperhidrosis (R2 = 0.0014, p = 9.55 × 10-3). Despite high genetic overlap between MDD and anxiety disorders, PRSs for anxiety disorders correlated with different types of sleep disturbances such as disorders of arousal or nightmares (R2 = 0.0013, p = 0.011). These findings suggest that greater genetic susceptibility to specific psychiatric and neurodevelopmental disorders, as represented by ADHD, MDD, and anxiety disorders, may contribute to greater sleep problems among children.

Project description:Three-quarters of lifetime mental illness occurs by the age of 24, but relatively little is known about how to robustly identify youth at risk to target intervention efforts known to improve outcomes. Barriers to knowledge have included obtaining robust predictions while simultaneously analyzing large numbers of different types of candidate predictors. In a new, large, transdiagnostic youth sample and multidomain high-dimension data, we used 160 candidate predictors encompassing neural, prenatal, developmental, physiologic, sociocultural, environmental, emotional and cognitive features and leveraged three different machine learning algorithms optimized with a novel artificial intelligence meta-learning technique to predict individual cases of anxiety, depression, attention deficit, disruptive behaviors and post-traumatic stress. Our models tested well in unseen, held-out data (AUC ≥ 0.94). By utilizing a large-scale design and advanced computational approaches, we were able to compare the relative predictive ability of neural versus psychosocial features in a principled manner and found that psychosocial features consistently outperformed neural metrics in their relative ability to deliver robust predictions of individual cases. We found that deep learning with artificial neural networks and tree-based learning with XGBoost outperformed logistic regression with ElasticNet, supporting the conceptualization of mental illnesses as multifactorial disease processes with non-linear relationships among predictors that can be robustly modeled with computational psychiatry techniques. To our knowledge, this is the first study to test the relative predictive ability of these gold-standard algorithms from different classes across multiple mental health conditions in youth within the same study design in multidomain data utilizing >100 candidate predictors. Further research is suggested to explore these findings in longitudinal data and validate results in an external dataset.