For decades, doctors have understood acute myeloid leukaemia (AML) mainly through its genetic mutations. Those mutations guide how the aggressive blood cancer is classified, how a patient's risk is judged and which drugs are chosen. Yet genetics has never fully explained why people with seemingly similar tumours can follow very different courses.
A large international study published in Nature on 8 July 2026 adds a complementary lens. Researchers led by Seishi Ogawa and Yotaro Ochi at Kyoto University, together with Sören Lehmann at the Karolinska Institute in Sweden, looked not at the DNA sequence itself but at how the genome is packaged — the chromatin state that determines which genes a cell can switch on.
Sixteen distinct profiles
Drawing on Swedish and Japanese patients, the team assembled the eCHROMA cohort of 1,563 people newly diagnosed with AML. Using ATAC-seq, a scalable method for reading chromatin accessibility, and combining it with data on gene mutations, gene activity, DNA methylation and histone marks, they found that AML sorts into 16 subgroups. Some map onto forms of the disease that are already recognised, while others emerged as new categories that conventional genetic classification had missed.
Each subgroup carried its own signature: characteristic mutations, differentiation state, patterns of gene expression and distinctive "super-enhancers" — regulatory switches driven by transcription factors central to normal blood formation. Single-cell analysis showed that every leukaemic cell within a subgroup shared the same chromatin fingerprint, and the groupings held up when tested against independent patient sets.
The practical promise lies in what this map adds. The chromatin-based subgroups predicted patient outcomes even after standard genetic risk was taken into account, and they were linked to particular drug sensitivities, some of them unexpected. Two patients carrying the same mutation might still fall into different epigenetic subgroups, with different outlooks and different treatment options.
The authors describe their atlas as a resource rather than a finished diagnostic tool. But it strengthens a hopeful shift in cancer medicine: leukaemia is not simply a disease of the genome. Reading the epigenome alongside it could, in time, support a more precise and individually tailored approach to a cancer that remains difficult to treat.