¹Semmelweis University, Department of Pharmacodynamics, Budapest, Hungary, ²Semmelweis University, NAP3.0-SE Neuropsychopharmacology Research Group, Hungarian Brain Research Program, Budapest, Hungary, ³Semmelweis University, Department of Psychiatry and Psychotherapy, and Department of Clinical Psychology, Budapest, Hungary

Background: Suicide attempt (SA) is a complex and heritable phenomenon posing a pressing issue in public health and a constant challenge in psychiatry. SA phenotype is strongly associated with psychiatric illnesses, particularly major depressive disorder, yet evidence from genetic and epidemiological studies suggests partially distinct underlying mechanisms. Polygenic risk scores (PRS) obtained from genome-wide association studies (GWAS) give a quantitative representation of the common genetic liability and provide a method to study the genetic relationships between suicide attempt and depression phenotypes.

Aims & Objectives: Our aim was to test PRS-s for SA calculated from large sample GWAS-s and the contribution of interactions with current depressive symptom as well as with early life stress or current stress to their predictive effect on SA in an independent sample. We also tested the effect of PRS-SA on lifetime history of depression.

Method: PRS-s for SA (PRS-SA) were constructed using GWAS summary statistics from International Psychiatric Genomics Consortium studies as discovery samples. In an independent target sample (NewMood) we analysed the effects of PRS-SA on history of suicide attempts and history of depression using logistic regression models. We also tested the effect of an interaction between PRS-SA and current depression symptoms, and between PRS-SA and early life stress or current stress on predicting history of sjuicide attempts. Performance of the models was evaluated through area under the curve (AUC), Akaike information criterion (AIC), and Nagelkerke R².

Results: Significant association was observed between PRS-SA and lifetime history of SA in the target sample. The AUC across models was between 0.54-0.57, Nagelkerke R² values were between 0.0047-0.0112, regression coefficients of PRS-SA showed p-values between 0.032-0.004. Multivariate logistic regression models including PRS, current depressive symptom severity, age, and sex showed strong overall model fit for SA history (Nagelkerke R²=0.266–0.270; likelihood-ratio χ²=260–264; p<10⁻54). Adding PRS to a model with symptom severity and covariates yielded only small improvement (Δχ²=3.50, p=0.061). No PRS x depressive symptom severity interaction was observed on predicting SA history (Δχ²=0.013, p=0.91; interaction p=0.86–0.99). In gene-environment interaction analyses, the environmental exposures displayed considerable main effects (OR 1.44–2.24, p<10⁻⁸), however, interaction between environmental effects and PRS-SA on SA history was not significant. PRS-SA was not significantly related to lifetime depression history (p 0.83–0.98; AUC 0.50-0.52).

Discussion & Conclusions: We found that PRS for suicide attempts significantly predicted history of suicide attempts but not history of depression in an independent cohort. Furthermore, no interaction effect in case of suicide attempt as an outcome was observed for PRS-SA and either depression symptom severity, or early life stress, or current life stress. Our findings support the presence of a measurable polygenic signal for suicide attempt, but call for further investigation regarding the genetic association between suicide and depression involving more diverse, broader, and sophisticated phenotypes. These findings contribute to ongoing efforts to characterise the genetic architecture of suicidal behaviour using independent samples and multiple phenotypes.