Abstract
BACKGROUND: Acute myeloid leukemia (AML) is a genetically heterogeneous disease in which glutamine (Gln) contributes to AML progression. Therefore, this study aimed to identify potential prognostic biomarkers for AML based on Gln metabolism-related genes. METHODS: Gln-related genes that were differentially expressed between Cancer Genome Atlas-based AML and normal samples were analyzed using the limma package. Univariate, least absolute shrinkage, selection operators, and stepwise Cox regression analyses were used to identify prognostic signatures. Risk score-based prognostic and nomogram models were constructed to predict the prognostic risk of AML. Subsequently, consistent cluster analysis was performed to stratify patients into different subtypes, and subtype-related module genes were screened using weighted gene co-expression network analysis. RESULTS: Through a series of regression analyses, HGF, ANGPTL3, MB, F2, CALR, EIF4EBP1, EPHX1, and PDHA1 were identified as potential prognostic biomarkers of AML. Prognostic and nomogram models constructed based on these genes could significantly differentiate between high- and low-risk AML with high predictive accuracy. The eight-signature also stratified patients with AML into two subtypes, among which Cluster 2 was prone to a high risk of AML prognosis. These two clusters exhibited different immune profiles. Of the subtype-related module genes, the HOXA and HOXB family genes may be genetic features of AML subtypes. CONCLUSION: Eight Gln metabolism-related genes were identified as potential biomarkers of AML to predict prognostic risk. The molecular subtypes clustered by these genes enabled prognostic risk stratification.