By using an unsupervised machine learning approach to look at genetic variation across the protein-coding genomes of 140,000 species, researchers at Harvard Medical School and Oxford University have developed a new variant classifying system that performed on a par with wet lab approaches.