Prognostic Prediction Of Cytogenetically Normal Acute Myeloid Leukemia Based On A Gene Expression Model

Background:Acute myeloid leukemia is a group of highly heterogeneous malignant hematological diseases.The European Leukemia Net(ELN),as a well-known clinical practice guideline,stratifies AML patients into three risk groups,based primarily on the detection of cytogenetic abnormalities.However,the prognosis of cytogenetically normal AML(CN-AML),which is the largest AML subset,can be hard to define.Moreover,the clinical outcomes associated with this subgroup are diverse.Purpose:In order to better assess the prognosis of CN-AML patients and develop a reliable treatment plan,we try to construct a multivariate Cox model that can accurately assess the prognosis of CN-AML patients.Methods:We collected five sets of transcriptome data of CN-AML patients from public databases.We first carried out a differential expression analysis by comparing bone marrow samples from CN-AML patients and healthy controls.Then we performed LASSO regression and stepwise regression analysis to confirm whether any combination of these differentially expressed genes(DEGs)could be used to accurately predict prognosis.Using these genes,we established a prognostic multivariate Cox model.Furthermore,we applied a log-rank test to compare the overall survival difference between the high and low-risk groups.And the time-dependent receiver operating characteristic(ROC)analysis was applied to calculate the area under the ROC curve(AUC)value at 1-,2-,3-years of the model.Results:The identified DEGs were enriched in immune-related pathways,which indicated that immune dysfunction may play an important role in the occurrence and development of CN-AML.Then,the BeatAML cohort,as the training dataset,was used to construct a prognostic Cox model named NEST(Nine-gEne SignaTure)model.The validity of NEST was confirmed in four external independent cohorts.Furthermore,the risk score predicted by the NEST model remained an independent prognostic factor in multivariate analyses,which was not affected by age or the presence of other gene mutations.Further analysis revealed that the NEST model was suitable for bone marrow mononuclear cell(BMMC)samples but not peripheral blood mononuclear cell(PBMC)samples,which indirectly indicated subtle differences between BMMCs and PBMCs.The fact that subtle differences exist in BMMC and PBMC suggests that we should pay more attention to the source of samples in future research.Conclusion:Our results validated the robustness and accuracy of the NEST model.And the fact that there are subtle differences between BMMC and PBMC suggests that future research should consider the sample origins more strictly.At the same time,our results indicate immune dysfunction might also play a vital role in the pathogenesis of CN-AML,which not only broadens our understanding of the "2-hit" leukemogenesis model,but also shed new light on the new potential therapeutic target.