Suicide, machine learning, and the brain

Suicide causes around 800,000 deaths worldwide each year, making premature deaths due to suicide a significant public health problem (WHO 2018). The expression of suicidal behaviors is a complex phenotype with underlying biological, psychological, clinical, and sociocultural risk factors (Turecki et al., 2019). From a neurobiological perspective, suicidal behaviors (ideation, attempt) are associated with the degree of anatomical, physiological, and neurotranscriptomic dysregulations of a network of brain regions. This network includes the anterior insula, anterior cingulate, and prefrontal cortical regions that mediate emotional-regulatory behaviors. From a clinical perspective, diagnostic measures of depressive symptoms in major depressive disorder and bipolar disorders (i.e., mood disorders) are associated with over sixty percent of suicide deaths (WHO 2018). Of interest, the brain anatomical and related depressive symptoms associated with suicidal behavior are characterized by regionally specific brain measures of aberrant gene expression profiles in the post-mortem brains of suicide completed donors (Jabbi et al., 2020). These convergent neurobiological correlates for depressive illness and suicidal phenotypes lead me to ask: 1) are regional brain gene expression dysregulations underlying depressive illness in mood disorders differentially associated with suicide mortality? 2) And crucially, how do the brain’s gene expression dysregulations measured in post-mortem studies relate to diagnostic and demographic variables to confer risk for suicide across the psychiatric continuum? Answering these questions using a deterministic post-mortem sample with well-characterized suicide outcomes could help identify the convergent molecular and clinical risk factors for suicide deaths across the lifespan. Furthermore, this line of research could resolve the specificity of how such clinically defined molecular risk factors for suicide mortality may differ across mood disorders and other mental illnesses, ultimately aiding in diagnostically informed suicide prevention. I reviewed the extant literature on studies of brain gene expression correlates of suicide by focusing on unique suicide-related gene expression dysregulations in mood disorders. Moving forward, I propose using machine learning tools to integrate brain-based gene expression measures in conjunction with clinical/demographic variables to identify molecular and phenotypic risk predictors for suicide mortality.