Prof. Jussi Jokinen

Dr. Jussi Jokinen earned his MD from the University of Helsinki, Finland and his PhD at the Karolinska Institutet, Stockholm, Sweden. He obtained a senior researcher position in 2009 at the Department of Clinical Neuroscience at the Karolinska Institutet financed by the Swedish Research Council. He became a Professor with tenure in the Department of Clinical Sciences, Psychiatry division in Umeå University in 2014. The major objective for his research is clinical suicide prevention. Several studies concern interrelationship between violence and suicide in different cohorts. He investigates risk factors for suicidal behaviors including neurobiological correlates to intermediate phenotypes in attempted suicide for treatment implications and prevention. Another major research focus is impulse control disorders and specifically compulsive sexual behavior disorder (CSBD), a new diagnosis in ICD 11. He has several current trials concerning biomarkers, treatment and outcomes of CSBD.

Suicidality phenotypes, consisting of suicidal ideation, suicide attempt, and suicide death, are all heritable but present unique challenges in biomarker studies due to their individual complexity, overlap with each other, and varying associations with psychiatric disorders. A substantial body of research shows that elevated inflammatory markers, stress system biology, neurotransmitters-related biomarkers, neurotrophic/neuroplasticity markers, lipid and metabolic biomarkers, dysregulation of the endocannabinoid system and genetic variants and epigenetic modifications are associated with suicidal behaviors. Further, impaired emotion regulation networks assessed with functional MRI and Connectivity patterns associated with impulsivity have been proposed to be related to suicidality. Because suicide is highly multifactorial, combinations of biomarkers—rather than single markers—are more likely to yield clinically meaningful predictions. AI enables pattern recognition beyond human capacity and is used in suicide prediction studies. Suicide‑related biomarker research often combines genomics, proteomics, metabolomics, neuroimaging, digital phenotyping, and clinical data. Because the data are high‑dimensional and heterogeneous, AI and machine‑learning methods can be useful.

Classical machine learning approaches (Random forests, Gradient boosting) can be used in biomarker prediction from omics/clinical data. Deep learning approaches (Neural networks on imaging, NLP models) have been used in neuroimaging data and speech models. Multimodal Models (GNNs, late fusion nets) apply to analyse data sets combining imaging, genetics and clinical data. Because clinical biomarker research requires high interpretability, Explainable AI (XAI) tools are widely used to identify which biomarkers or features contribute most to risk prediction. Current predictive performance in research settings have reported Areas under the curve (AUCs) between 0.7–0.9. However, there are many challenges. AI‑driven suicide biomarker prediction faces significant challenges related to biomarker validity, data quality, clinical generalizability, ethical risks, and the complexity of suicidality as a phenomenon. Advanced ML models (e.g., neural networks, deep learning) often act as "black boxes", limiting clinicians’ ability to understand why a model flags an individual as high‑risk. Suicide results from complex interactions among biological, psychological, social, and environmental factors. Even the most accurate models struggle to incorporate this multifactorial dynamic, resulting in limited predictive precision. Neurobiological models alone cannot account for social stressors, trauma, substance use, and acute crisis periods—making comprehensive AI prediction very challenging.

• Jokinen J, et al. (2007). DST non-suppression predicts suicide after attempted suicide. Psychiatry Research, 150(3), 297–303. doi: 10.1016/j.psychres.2006.12.001.
• Isung J, <...> & Jokinen J. (2012). Low vascular endothelial growth factor and interleukin-8 in cerebrospinal fluid of suicide attempters. Translational Psychiatry, 2, e196. doi: 10.1038/tp.2012.123.
• Jokinen J, et al. (2018). Epigenetic changes in the CRH gene are related to severity of suicide attempt and a general psychiatric risk score in adolescents. EBioMedicine, 27, 123–133. doi: 10.1016/j.ebiom.2017.12.018.
• Chen Q, Zhang-James Y, Barnett EJ, Lichtenstein P, Jokinen J, et al. (2020). Predicting suicide attempt or suicide death following a visit to psychiatric specialty care: A machine learning study using Swedish national registry data. PLOS Medicine, 17(11), e1003416. doi: 10.1371/journal.pmed.1003416.