Genomic Exploration of Brain Structure and Function and Behavioral Health in the Context of Suicidal Behaviors

Suicidal behaviors are a major public health problem with a substantial impact on individuals, families, and society. Different approaches ranging from genetic studies to influencing factors analyses and predictive modeling have been explored to understand mechanisms leading to suicide ideation and attempt. However, the previous studies have limitations. For instance, regression models were mainly applied to observational data to assess suicide-related associations. In suicide predictive modeling, only few widely used algorithms were explored and the performance reported appeared to have limited applicability to real-world scenarios. Similarly, large-scale genome-wide association studies (GWAS) identified many risk loci, but still did not provide a clear understanding of which brain mechanisms can increase the risk of suicidal behaviors. The proposed study aims to expand the comprehension of the genetic basis of suicide ideation and attempt and integrate them with molecular data derived from brain tissues and imaging information regarding brain structure and function. Specifically, we will use genome-wide data from six cohorts, including a total of suicidal-ideation 122,310 cases and 46,332 suicidal-attempt cases. Among the cohorts investigated, we will analyze novel data from the All of Us Research Program, which will permit us to generate information regarding the genetics of suicidal behaviors in human groups reflecting the diversity of the US population. The genome-wide associations will be integrated with brain multi-omics data to perform transcriptome-wide association and proteome-wide association studies to uncover the molecular dynamics leading to suicide ideation and attempt. Additionally, we will use the newly generated genetic data to investigate possible direct and interactive effects linking suicidal behaviors to brain structure and function, behavioral health, and other clinically relevant outcomes using multiple approaches (i.e., phenome-wide association analyses, polygenic risk scoring, Mendelian randomization, and gene-by-environment genome-wide interaction investigations). Based on results identified from molecular, imaging, and clinical data, we will use a cutting-edge framework applying a multi-model strategy to develop a multi-omics and multi-trait model to effectively predict suicide ideation and attempt. The integration of multiple data types and multiple machine learning algorithms is expected to boost the prediction power of the algorithms that will be developed. Additionally, to unveil the “black box” in machine learning, we will further investigate the best predictive model by assessing the contribution of the features and their relationship with suicidal behaviors. The expected findings of the proposed study will contribute to disentangle the biology and the epidemiology of suicidal behaviors and translate the results into instruments that could guide preventive strategies, facilitating the early identification of individuals at high suicide risk.