Table_5_A Prognostic Model for Acute Myeloid Leukemia Based on IL-2/STAT5 Pathway-Related Genes.xlsx

Accurate prognostic stratification of patients can provide guidance for personalized therapy. Many prognostic models for acute myeloid leukemia (AML) have been reported, but most have considerable inaccuracies due to contained variables with insufficient capacity of predicting survival and lack of adequate verification. Here, 235 genes strongly related to survival in AML were systematically identified through univariate Cox regression analysis of eight independent AML datasets. Pathway enrichment analysis of these 235 genes revealed that the IL-2/STAT5 signaling pathway was the most highly enriched. Through Cox proportional-hazards regression model and stepwise algorithm, we constructed a six-gene STAT5-associated signature based on the most robustly survival-related genes related to the IL-2/STAT5 signaling pathway. Good prognostic performance was observed in the training cohort (GSE37642-GPL96), and the signature was validated in seven other validation cohorts. As an independent prognostic factor, the STAT5-associated signature was positively correlated with patient age and ELN2017 risk levels. An integrated score based on these three prognostic factors had higher prognostic accuracy than the ELN2017 risk category. Characterization of immune cell infiltration indicated that impaired B-cell adaptive immunity, immunosuppressive effects, serious infection, and weakened anti-inflammatory function tended to accompany high-risk patients. Analysis of in-house clinical samples revealed that the STAT5-assocaited signature risk scores of AML patients were significantly higher than those of healthy people. Five chemotherapeutic drugs that were effective in these high-risk patients were screened in silico. Among the five drugs, MS.275, a known HDAC inhibitor, selectively suppressed the proliferation of cancer cells with high STAT5 phosphorylation levels in vitro. Taken together, the data indicate that the STAT5-associated signature is a reliable prognostic model that can be used to optimize prognostic stratification and guide personalized AML treatments.

精确的预后分层对于个性化治疗具有指导意义。目前已有许多关于急性髓系白血病（AML）的预后模型被报道，但大多数模型由于包含变量不足以及预测生存能力的有限，且缺乏充分的验证，因而存在较大的不准确度。在本研究中，通过分析八个独立AML数据集的单变量Cox回归分析，系统地识别出与AML生存高度相关的235个基因。对这些235个基因进行通路富集分析发现，IL-2/STAT5信号通路富集程度最高。通过Cox比例风险回归模型和逐步算法，我们基于与IL-2/STAT5信号通路相关的最稳健的生存相关基因构建了一个包含六个基因的STAT5相关特征。在训练队列（GSE37642-GPL96）中观察到良好的预后性能，且该特征在七个其他验证队列中得到验证。作为独立的预后因素，STAT5相关特征与患者年龄和ELN2017风险等级呈正相关。基于这三个预后因素的集成评分比ELN2017风险类别具有更高的预后准确性。对免疫细胞浸润的分析表明，B细胞适应性免疫受损、免疫抑制效应、严重感染以及抗炎功能减弱往往伴随高风险患者。对内部临床样本的分析揭示，AML患者的STAT5相关特征风险评分显著高于健康人群。通过虚拟筛选，筛选出五种对高风险患者有效的化疗药物。在这五种药物中，MS.275，一种已知的HDAC抑制剂，在体外选择性地抑制了STAT5磷酸化水平高的癌细胞增殖。综合以上数据，表明STAT5相关特征是一个可靠的预后模型，可用于优化预后分层并指导AML的个性化治疗。