Combined family history and polygenic score prediction of major depressive disorder
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Importance
Major depressive disorder (MDD) is a complex psychiatric disorder influenced by genetic, social, and environmental factors. Family history, genome-wide polygenic scores, and childhood trauma are key predictors of MDD onset with distinct contributions.
Objective
This study modelled the combined effects of family history of multiple psychiatric disorders, polygenic scores of multiple traits, and childhood trauma, alongside sociodemographic factors on MDD diagnosis and number of episodes. We aimed to build and externally validate predictive models for MDD risk and severity.
Participants
We used data from the Genetic Links to Anxiety and Depression Study and other NIHR BioResource studies (GLAD+) and UK Biobank (UKB) collected between 2016 and 2023.
Outcomes and measurements
MDD diagnosis followed DSM-5 criteria using online questionnaire data. Family history (Yes/No) was reported for up to 22 psychiatric disorders. Polygenic scores were calculated based on genome-wide association studies (n=22). Participants answered the five-item childhood trauma screener. We used elastic net regression with nested cross-validation to select the best predictors.
Results
In GLAD+ (9,927 MDD cases, 4,452 controls), family history explained 17% of the MDD variance, followed by childhood trauma (11%), sociodemographics (10%), and polygenic scores (7%), resulting in 33% of the MDD variance (AUC-ROC=0.84). In UKB (40,667 MDD cases, 70,755 controls), family history explained 13% of the MDD variance, childhood trauma (7%), sociodemographics (6%), and polygenic scores (4%). Combined together, the predictors explained 23% of the variance (AUC-ROC=0.74). The top five individual predictors were family history of depression, childhood trauma, female sex, family history of anxiety, and the MDD polygenic score in both cohorts. The predictive models were externally well-validated across GLAD+ and UKB obtaining comparable predictive performance. Finally, when combined, the predictors explained 26% and 12% of the variance in the number of MDD episodes in GLAD+ and UKB respectively.
Conclusions
Integrating family history, PRSs, ChT, and sociodemographic factors can predict risk of an MDD diagnosis and its recurrent course. This model may help identify individuals at high risk for depression in clinical settings, assess its severity, enable early diagnosis and potentially personalize treatment.
