A novel prognostic model of methylation-associated genes in acute myeloid leukemia.

Background: There is growing evidence that methylation-associated genes (MAGs) play an important role in the prognosis of acute myeloid leukemia (AML) patients. Thus, the aim of this research was to investigate the impact of MAGs in predicting the outcomes of AML patients.

Methods: The expression profile and clinical information of patients were downloaded from public databases. A novel prognostic model based on 7 MAGs was established in the TCGA training cohort and validated in the GSE71014 dataset. To validate the clinical implications, the correlation between MAGs signature and drug sensitivity was further investigated.

Results: 76 genes were screened out by the univariate Cox regression and significantly enriched in multiple methylation-related pathways. After filtering variables using LASSO regression analysis, 7 MAGs were introduced to construct the predictive model. The survival analysis showed overall survival of patients with the high-risk score was considerably poorer than that with the low-risk score in both the training and validating cohorts (p < 0.01). Furthermore, the risk score system as a prognostic factor also worked in the intermediate-risk patients based on ELN-2017 classification. Importantly, the risk score was demonstrated to be an independent prognostic factor for AML in the univariate and multivariate Cox regression analysis. Interestingly, GSEA analysis revealed that multiple metabolism-related pathways were significantly enriched in the high-risk group. Drug sensitivity analysis showed there was a significant difference in sensitivity of some drugs between the two groups.

Conclusion: We developed a robust and accurate prognostic model with 7 MAGs. Our findings might provide a reference for the clinical prognosis and management of AML.