The current study suggests that alcohol-related memory problems can be predicted using a multi-domain set of features from neural, behavioral, genomic, and alcohol-related measures in a machine-learning framework. It was found that the memory group showed a predominant pattern of hyperconnectivity across the default mode network regions, including the hippocampal subnetworks, while showing hypoconnected anterior cingulate cortex subnetworks; these results were based on the EEG recorded about 18 years ago. Features from other domains that significantly contributed to the classification were (i) higher counts of alcohol-related consequences during the past five years, such as health problems, other alcohol-related adverse past negative experiences, withdrawal symptoms, and a higher max number of drinks (the largest number of drinks per day), (iii) personality factors such as high neuroticism, high harm avoidance, and low rates of positive/uplifting experiences, and (iv) high genetic liability, as reflected in variations in PRS for AUD across the memory and control groups. It should also be noted that the classification accuracy was better for the control individuals (85/94 = 90.43%) than for the memory group (68/94 = 72.34%). Although
