Abstract:Current MDS and AML classifications are based on a hierarchical interplay of morphology (green) and genetics (blue). In future, disease classification could be based primarily on unbiased clustering analysis of genetic aberrations, gene expression and methylation profile; morphology and ontogeny would be separate qualifiers that further inform therapeutic decisionmaking. Myelodysplastic syndromes/neoplasms (MDS) and acute myeloid leukaemia (AML) are neoplastic haematopoietic cell proliferations that are diagnosed and classified based on a combination of morphological, clinical and genetic features. Specifically, the percentage of myeloblasts in the blood and bone marrow is a key feature that has historically separated MDS from AML and, together with several other morphological parameters, defines distinct disease entities within MDS. Both MDS and AML have recurrent genetic abnormalities that are increasingly influencing their definitions and subclassification. For example, in 2022, two new MDS entities were recognised based on the presence of SF3B1 mutation or bi‐allelic TP53 abnormalities. Genomic information is more objective and reproducible than morphological analyses, which are subject to interobserver variability and arbitrary numeric cut‐offs. Nevertheless, the integration of genomic data with traditional morphological features in myeloid neoplasm classification has proved challenging by virtue of its sheer complexity; gene expression and methylation profiling also can provide information regarding disease pathogenesis, adding to the complexity. New machine‐learning technologies have the potential to effectively integrate multiple diagnostic modalities and improve on historical classification systems. Going forward, the application of machine learning and advanced statistical methods to large patient cohorts can refine future classifications by advancing unbiased and robust previously unrecognised disease subgroups. Future classifications will probably incorporate these newer technologies and higher‐level analyses that emphasise genomic disease entities over traditional morphologically defined entities, thus promoting more accurate diagnosis and patient risk stratification.

Future directions in myelodysplastic syndromes/neoplasms and acute myeloid leukaemia classification: from blast counts to biology