Construction of a Stromal-Related Prognostic Model in Acute Myeloid Leukemia by Comprehensive Bioinformatics Analysis Publisher



Hajipirloo LK¹ ; Nabigol M¹ ; Khayami R² ; Karami N³ ; Farsani MA¹ ; Navidinia AA^{1, 3}
Source: Current Research in Translational Medicine Published:2025

Abstract

Background: Stromal cells play a pivotal role in the tumor microenvironment (TME), significantly impacting the progression of acute myeloid leukemia (AML). This study sought to develop a stromal-related prognostic model for AML, aiming to uncover novel prognostic markers and therapeutic targets. Methods: RNA expression data and clinical profiles of AML patients were retrieved from the Cancer Genome Atlas (TCGA). The extent of stromal cell infiltration within the TME was quantified using the ESTIMATE algorithm. Associations between stromal scores and the French-American-British (FAB) classification, overall survival (OS), and the Cancer and Leukemia Group B (CALGB) cytogenetic risk categories were analyzed. Differentially expressed genes (DEGs) were identified, and gene ontology (GO) and protein-protein interaction (PPI) networks were constructed. Prognostic DEGs were selected through LASSO-cox regression analysis. A risk score model was then developed based on these DEGs. A stromal-related prognostic model (SPM) was constructed from the patients' risk scores (RS), and its efficacy was evaluated using Receiver Operating Characteristic (ROC) curves and a nomogram. The association between FAB, CALGB, age, and common mutations and SPM was also assessed. Ultimately, the SPM was validated using an external dataset from 246 patients in the TARGET-AML study. Results: Kaplan-Meier analysis revealed a significant association between stromal scores and patient survival (p = 0.04). LASSO[sbnd]Cox regression identified four genes (MAP7D2, CDRT1, HOXB9, and IRX5) as highly predictive of survival. The prognostic model showed a strong correlation with overall survival, with higher scores indicating poorer outcomes (p = 1.48e-07). Older patients (over 60 years) faced significantly worse prognoses (p = 0.0055). Although no significant association was found between the SPM and the FAB classification (p = 0.063), both poor and intermediate/normal cytogenetic groups had significantly higher SPM risk scores than the favorable group (p = 0.0057 and 0.0026). External validation of the SPM in the TARGET-AML dataset confirmed a significant association with survival (p = 0.00035), with the area under the curve (AUC) for 10-year survival at 75.81 %. Conclusion: Our research successfully established a stromal-related prognostic model in AML, offering new perspectives for prognostic evaluation and identifying potential targets for therapeutic intervention. © 2025 Elsevier Masson SAS