Genomic lesions including somatic mutations and translocations from >6000 AML patients were aggregated. Machine learning identified 4 underlying genomic clusters based on molecular features that were then shown to have differing survival rates, implying underlying differences in biology.