Abstract
Acute Myeloid Leukemia (AML) is a genetically and clinically heterogeneous disease that can develop at any age. While AML incidence increases with age and distinct genetic alterations are observed in younger versus older patients, current classification systems do not incorporate age as a defining factor. In this study, we analyzed RNA-seq data from 404 AML patients at initial diagnosis, leveraging a k-mer-based machine learning approach to uncover age-related transcriptomic differences in favorable and adverse risk groups. Our model achieved over 90% accuracy in risk prediction and identified key gene signatures distinguishing ELN2017 favorable and adverse groups. From these signatures, we selected prognostic biomarkers with significant impacts on survival. Additionally, we explored the biological context underlying transcriptomic complexity across age groups, revealing distinct tumor profiles and differences in immune and stromal cell populations, particularly in older patients. These findings underscore the importance of age-related molecular features in AML and provide new insights for risk stratification and therapeutic targeting.