Transcriptome-based molecular subgroup identification and prognosis stratification in pediatric AML.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
Current risk subgroups for acute myeloid leukemia (AML) rely primarily on molecular and cytogenetic markers. These subgroups have distinct clinicopathological and molecular characteristics and are the basis for classifying disease subgroups currently undergoing clinical trials and initial subgroup-oriented therapies. However, substantial biological heterogeneity and differences in survival are apparent within each subgroup, so transcriptome-based molecular subgroup identification and prognostic stratification in pediatric acute myeloid leukemia remain to be resolved. In this study, we conducted a comprehensive analysis of transcriptomic data for AML using data from public datasets and constructed a new prognostic prediction model. We first downloaded the transcriptome data of the GDC Data Portal and the gene set of MsigDB for survival and cluster analysis and reclassified AML into 8 different molecular subgroups. The expression profiles, changes in the immune microenvironment, biological functions and pathways, and clinical features among these subpopulations were further studied. The classification prediction model is then developed using four machine learning algorithms: RF, SVM, XGBoost, and DT. The XGBoost method showed the best performance. The vital feature variables in XGBoost suggest that HSD17B10, NDUFS8, ASCL5, FADS2, and COX8A were critical factors in identifying the prognosis of AML. We only found that NDUFS8 and FADS2 had been reported in AML, and the remaining three gene new prognostic markers for AML have not been reported before, but they have all been reported to be associated with cancer. we identified eight transcriptome-based molecular subgroups of AML and further assessed differences in the molecular landscape, immune networks, and signaling pathways underlying these subpopulations. Additionally, the prognostic models comprising 62 genes was proposed and demonstrated to have remarkable predictive value. Overall, we identified a new molecular subpopulation of AML, which improved disease risk stratification and established a prognostic model for AML, which demonstrated favorable performance in retrospective validation. Since this study relies on retrospective data, further prospective analysis is required to confirm its ability to accurately reflect patient prognosis.