Background: Acute Myeloid Leukemia (AML) is characterized as a type of hematological malignancy with poor survival. Accumulated evidence showed that dysregulated immune activities contribute to the pathogenesis of AML and accelerate the development of chemotherapy resistance. Thus, we aimed to construct prognostic signatures based on patients’ immune features to sort out the high-risk group and to identify survival-related checkpoint molecules as potential therapeutic targets.Methods: In the current study, we developed two prognostic signatures based on immune genes and infiltrated fraction of immune cells, respectively, using a least absolute shrinkage and selection operator model, and Cox regression analysis on 415 samples obtained from TCGA and GEO databases. Results: We found the optimum strategy for predicting patients’ survival is combined using these two prognostic immune-related signatures. Through our established signatures, we classified patients into Favorable Risk group and Poor Risk group, who showed significantly different OS and DFS. We further demonstrated the checkpoint molecules’ profile in different risk groups. Conclusions: we constructed a powerful prognostic tool here to help classify high-risk patients in early-stage, who may benefit from additional immune therapies by targeting identified checkpoint molecules.