RESUMEN
Background: Acute Myeloid Leukemia (AML) exhibits a wide array of phenotypic manifestations, progression patterns, and heterogeneous responses to immunotherapies, suggesting involvement of complex immunobiological mechanisms. This investigation aimed to develop an integrated prognostic model for AML by incorporating cancer driver genes, along with clinical and phenotypic characteristics of the disease, and to assess its implications for immunotherapy responsiveness. Methods: Critical oncogenic driver genes linked to survival were identified by screening primary effector and corresponding gene pairs using data from The Cancer Genome Atlas (TCGA), through univariate Cox proportional hazard regression analysis. This was independently verified using dataset GSE37642. Primary effector genes were further refined using LASSO regression. Transcriptomic profiling was quantified using multivariate Cox regression, and the derived prognostic score was subsequently validated. Finally, a multivariate Cox regression model was developed, incorporating the transcriptomic score along with clinical parameters such as age, gender, and French-American-British (FAB) classification subtype. The 'Accurate Prediction Model of AML Overall Survival Score' (APMAO) was developed and subsequently validated. Investigations were conducted into functional pathway enrichment, alterations in the gene mutational landscape, and the extent of immune cell infiltration associated with varying APMAO scores. To further investigate the potential of APMAO scores as a predictive biomarker for responsiveness to cancer immunotherapy, we conducted a series of analyses. These included examining the expression profiles of genes related to immune checkpoints, the interferon-gamma signaling pathway, and m6A regulation. Additionally, we explored the relationship between these gene expression patterns and the Tumor Immune Dysfunction and Exclusion (TIDE) dysfunction scores. Results: Through the screening of 95 cancer genes associated with survival and 313 interacting gene pairs, seven genes (ACSL6, MAP3K1, CHIC2, HIP1, PTPN6, TFEB, and DAXX) were identified, leading to the derivation of a transcriptional score. Age and the transcriptional score were significant predictors in Cox regression analysis and were integral to the development of the final APMAO model, which exhibited an AUC greater than 0.75 and was successfully validated. Notable differences were observed in the distribution of the transcriptional score, age, cytogenetic risk categories, and French-American-British (FAB) classification between high and low APMAO groups. Samples with high APMAO scores demonstrated significantly higher mutation rates and pathway enrichments in NFKB, TNF, JAK-STAT, and NOTCH signaling. Additionally, variations in immune cell infiltration and immune checkpoint expression, activation of the interferon-γ pathway, and expression of m6A regulators were noted, including a negative correlation between CD160, m6A expression, and APMAO scores. Conclusion: The combined APMAO score integrating transcriptional and clinical parameters demonstrated robust prognostic performance in predicting AML survival outcomes. It was linked to unique phenotypic characteristics, distinctive immune and mutational profiles, and patterns of expression for markers related to immunotherapy sensitivity. These observations suggest the potential for facilitating precision immunotherapy and advocate for its exploration in upcoming clinical trials.