Abstract
Acute myeloid leukemia (AML) is a malignancy of the stem cell precursors of the myeloid lineage. DNA methylation is an important DNA modification that regulates gene expression. Investigating AML heterogeneity based on DNA methylation could be clinically informative for improving clinical diagnosis and prognosis. The AML subtypes based on DNA methylation were identified by unsupervised consensus clustering. The association of these subtypes with gene mutation, copy number variations, immune infiltration and clinical features were further explored. Finally, univariate, LASSO and multivariate cox regression analyses were used to identify prognosis-associated genes and construct risk model for AML patients. In addition, we validated this model by using other datasets and explored the involved biological functions and pathways of its related genes. Three CpG island methylator phenotypes (CIMP-H, CIMP-M and CIMP-L) were identified using the 91 differential CpG sites. Overall survival, morphology, macrophages M0 and monocytes were distinct from each other. The most frequently mutated gene in CIMP-L was DNMT3A while which in CIMP-M that was RUNX1. In addition, the TIDE scores, used to predict the response to immune checkpoint inhibitors, were significantly different among CIMPs. The CIMP-associated prognosis risk model (CPM) using 32 key genes had convinced accuracy of prediction to forecast 0.5-year, 1-year, 3-year and 5-year survival rates. Moreover, the risk score-related genes were significantly enriched in pattern specification process, regionalization, embryonic organ morphogenesis and other critical cancer-related biological functions. We systematically and comprehensively analyzed the DNA methylation in AML. The risk model we constructed is an independent predictor of overall survival in AML and could be used as prognostic factor for AML treatment.