Given the uniqueness of the COGA dataset (with genetics, EEG measurements, and AUD remitter status), analysis on an independent dataset was not available. The latest prediction models’ approach is towards precision medicine, in which sex and ancestral stratification analysis produce more group-specific tailored results. This approach led to different sample sizes, with AA groups showing a smaller cohort. For homogeneous analysis across different group’ sizes, we applied CV analysis on all models, while additional training/testing validation was applied and confirmed the CV results on the larger samples (EA males and females, p > 0.1 for all models). Future studies with larger cohorts are required to further validate these results. Another limitation is related to the scope of features. Various symptomatic and psychosocial features were implicated in previous studies as associated with AUD development, including our own work8. These features were not included in the current analysis to enable a focus on biomarkers (genetics, brain function) for prediction. Future studies with a wider selection of features are required to further investigate the variables that best predict remission from AUD.
