Identifying who is vulnerable to develop Alcohol Use Disorder (AUD), determining ‘sensitive periods’, and finding relevant biological markers are a major challenge. Studies show that rates of alcohol consumption dramatically increase during the teenage years 1 and genetic and environmental factors can increase the risk for transitioning to AUD 2. However, clear indication as to who is prone to develop AUD is still unclear. Recent studies have suggested that multidimensional modeling of genetic, biological, and psychosocial information may better reflect the underlying pathophysiology compared to one-dimensional measures 3, 4. Indeed, over the last decade, machine learning (ML) approaches and data mining processes have been successfully applied for analysis of multidimensional datasets including neuroimaging and genetic data to help in the context of disease diagnosis 5, 6, outperforming classical regression approaches 7. ML Support Vector Machine (SVM) classifiers have succeeded in predicting diagnosis, clinical outcomes, and classifying disorders such as depression 6, schizophrenia 4, 8, and AUD 9–11. Specifically, AUD classifiers achieved significant accuracy utilizing electrophysiological features such as EEG coherence and spectral power (89.3%)10, 11, EEG’s nonlinear features (91.7%) 9,
