Refining AML patients stratification is a major challenge and represents one of the keystones of both diagnosis and therapy improvement. As part of the Leucegene project, our group sequenced close to 500 primary AML specimens (DNA and/or RNA) selected to best represent the genetic diversity of the disease. Using this unique dataset, we are developing multidisciplinary projects combining -omics, molecular biology and innovative analysis methods, including machine learning, to dissect the determinants that shape AML variability (e.g. mutations, structural rearrangements, deferentially expressed genes). More specifically, our approach involves the comparison of data from specimens of a given genetic subgroup with sequencing results from the other specimens of the cohort, with a focus on gene/mutation signatures comprising a limited number of candidates, rather than on more conventional global clustering. Using this strategy, we confirmed all known subgroup-specific mutations in AML, and more importantly, we identified new ones allowing patient reclassification. Leveraging our unique dataset, we aim to better characterize (rare) AML subgroups and identify new ones, to ultimately propose new diagnostic and prognostic markers.