discriminative key features for every group. To further increase group ancestry genetic homogeneity we calculated the models when ancestry was based on self-report and again based on genetics calculated with ancestral principal components (PCA)21. Our central hypothesis was that model based on multidimensional features will result in a better prediction than singular modality and that being married, employed, and taking medication will predict remission. Using stratification to control for the confounding variables, sex, and ancestry, we expected to find differences in the prediction models between the groups, with higher accuracy scores when the ancestry was calculated using genetic data. We also examined the most discriminative features in the predictive models, enhancing our understanding of neurophysiological, genetic, and socio-demographic characteristics underlying AUD resilience and recovery.
