〔Abstract〕 To investigate the aberrant alterations of microRNAs ( miRNAs) in exosomes derived from bone marrow stromal cells ( BMSCs) in the bone marrow supernatants of patients with acute myeloid leukemia (AML) and their impact on the prognosis of AML patients . Methods Bone marrow supernatant samples were col- lected from three AML patients and three healthy donors . Exosomes were isolated using a commercial kit , identif- ying the morphology and marker expression , and subjected to miRNA sequencing to determine differentially ex- pressed miRNAs (DE-miRNAs) . The DE-miRNAs were then intersected with the exosomal miRNA expression pro- files of primary AML cells (GSE64029) to exclude AML cell — derived signals and to identify BMSC-derived DE - miRNAs . Subsequently , candidate miRNAs were identified through Cox regression and Lasso regression analyses based on data from The Cancer Genome Atlas (TCGA) . A prognostic risk model for AML was constructed , and pa- tients were stratified into high-risk and low-risk groups according to the median risk score . The prognostic value and clinical relevance of the model were further validated . Finally , the target genes of the candidate miRNAs were pre- dicted , followed by pathway enrichment analysis , construction of key regulatory networks , and correlation analysis between the expression levels of key miRNAs and their corresponding target genes . Results Isolated exosomes ex- hibited a typical cup-shaped morphology with intact structures with particle size of 30 - 150 nm , and expressed exo- somal markers CD63 , ALIX , and TSG101 . miRNA sequencing identified 103 DE-miRNAs in AML patients com- pared with healthy donors; after intersection with the GSE64029 dataset , 83 BMSC-derived DE-miRNAs were re- tained . Among these , five candidate miRNAs ( miR-25-3p , miR-532-5p , miR-194-5p , miR-10a-5p , and miR- 20a-5p) were used to construct the prognostic model . Kaplan — Meier survival analysis demonstrated significantly longer overall survival in the low-risk group compared with the high-risk group ( P < 0. 05) . The areas under the ROC curve for the training/validation cohorts were 0. 80/0. 74 , 0. 80/0. 78 , and 0. 79/0. 64 at 1 , 2 , and 3 years , respectively . The prognostic model was significantly associated with risk stratification , patient age , and FAB classi- fication (P < 0. 05) . KEGG pathway enrichment revealed that target genes of the candidate miRNAs were closely linked to cancer-related signaling pathways , including hepatocellular carcinoma , breast cancer , and non-small cell lung cancer. Correlation analysis indicated that the candidate miRNAs were significantly associated with key genes such as HIF1A , CREB1 , PIK3CA , IGF1R , PIK3R1 , TIAM1 , CRK , and PTEN (P < 0. 05) . Conclusion AML patients exhibit distinct miRNA expression profiles in BMSC-derived exosomes . A five-miRNA signature ( miR-25 - 3p , miR-532-5p , miR-194-5p , miR-10a-5p , and miR-20a-5p) demonstrates robust prognostic performance , sup- porting its potential clinical utility in risk stratification and outcome prediction for AML.