We propose a state-of-the-art, scalable and flexible alternative to the classical GWAS approach, based on machine learning techniques, to analyze large-scale data and discover epistatic and non-epistatic polygenic variants ...