Biomarkers for Risk of Phenotype Suicide Behaviors

Abstract

The purpose of this study was to determine if methylation of blood DNA predicted whether an individual would attempt suicide. It was not known if blood contained epigenetic biomarkers that were reliable predictors of attempting suicide. This work was an evaluation of quantitative methods and machine learning models in a correlational predictive design. The first step of the study was an evaluation to determine to what extent measurements of the methylation of blood DNA contributed to accurately predicting whether an individual had attempted suicide. The second step was an evaluation to determine if an association existed between those methylations that supported acceptable predictions and life stresses the individual had experienced. Archival Emory University data describing 488 urban, low income, African American women recruited from clinics of Grady Memorial Hospital in Atlanta, Georgia were the source for the study. The results of the study were that some models supported predictions of suicide attempts at AUC greater than .7 but no model had the sensitivity and specificity necessary for clinical use. Other results of the study were that five methylations that contributed to the models were also associated with life stress. The models in the study provided support for the involvement in psychological disorders and suicidality of 56 genes identified by several previous studies, eight genes that one previous study identified, and eleven novel genes.