This study modelled the latent genetic structure of twelve psychiatric and personality phenotypes using summary statistics from large European-ancestry GWAS. The authors report that a three-factor solution provided the best fit. The three latent factors were characterized as follows: F1 (post-traumatic stress disorder, major depressive disorder, generalized anxiety disorder, attention-deficit/hyperactivity disorder, and agreeableness), F2 (bipolar disorder, schizophrenia, obsessive-compulsive disorder, and openness), and F3 (obsessive-compulsive disorder, generalized anxiety disorder, extraversion, and neuroticism).
The modelling approach—Genomic Structural Equation Modelling (Genomic SEM)—was used to estimate the shared genetic architecture across these traits, enabling the identification of latent dimensions that capture genetic covariance among psychiatric disorders and personality measures.
The authors examined how the identified latent factors correlated genetically with six substance use behaviors. Across analyses, smoking-related phenotypes showed significant genetic correlations with each latent factor, with the strongest correlation observed between smoking measures and F1. Alcohol use exhibited weak but statistically significant correlations with F1 and F2. Cannabis ever-use demonstrated moderate correlations with F1 and F2, and the strongest genetic correlation of cannabis use was with F2.
These pattern differences indicate that distinct combinations of psychiatric and personality genetic liabilities relate differentially to specific substance use behaviors, suggesting partially separable genetic pathways linking psychopathology, personality traits, and substance use.
To probe potential causal directions underlying associations, the study applied bidirectional Mendelian Randomization (MR) analyses between substance use phenotypes and psychiatric/personality traits. The MR results reported several relatively robust causal relationships:
The authors note that many other tested associations provided only weak causal evidence. For a number of analyses, MR estimates were affected by heterogeneity and pleiotropy, which complicates causal interpretation and reduces confidence in those specific findings.
All analyses used previously published GWAS summary statistics drawn from multiple consortia and studies. The manuscript and supplementary materials list the specific data sources and provide links to the summary statistics used for substance use traits (age of smoking initiation, cigarettes per day, drinks per week, smoking cessation, smoking initiation, cannabis ever use) and for psychiatric and personality GWAS (agreeableness, extraversion, neuroticism, conscientiousness, openness; ADHD; bipolar disorder; generalized anxiety; major depression; obsessive-compulsive disorder; post-traumatic stress disorder; schizophrenia). The authors state that all necessary ethical approvals and participant consent processes were followed for the original GWAS, and that no new individual-level participant data were collected for this study. The work is provided as a preprint; it has not been certified by peer review.
The authors highlight key limitations relevant to interpretation. Several MR associations showed weak evidence or were affected by instrument heterogeneity and pleiotropy, which limit firm causal claims for many tested relationships. Because the analyses relied on summary-level GWAS data from participants of European ancestry, generalizability to other ancestry groups is not addressed in the report. Additional limitations inherent to Genomic SEM and MR—such as reliance on the quality of the underlying GWAS, potential measurement heterogeneity across contributing studies, and the assumption sets required for causal inference by MR—are implicit in the findings.
The preprint status of the manuscript is important: these results have not been peer reviewed, and the authors caution against using the findings to guide clinical practice at this stage.
Using complementary genetically informed methods, the study provides evidence for both shared genetic liability and a subset of direct causal effects linking substance use behaviors, psychiatric disorders, and personality traits. Distinct latent genetic factors showed differential correlations with smoking, alcohol, and cannabis outcomes, and MR analyses suggested specific causal pathways (for example, smoking-related measures to bipolar disorder and schizophrenia, and schizophrenia/PTSD liability to cannabis use). Many associations remain uncertain due to heterogeneity and pleiotropy.
These results aim to inform future prevention and intervention research by clarifying which relationships are more likely to arise from shared genetic liability versus potentially causal genetic pathways. The manuscript includes links to all GWAS sources and supplementary materials for researchers who wish to examine the analyses in detail.