Accumulating evidence revealed that autophagy played vital roles in breast cancer (BC) progression. Thus, the aim of this study was to investigate the prognostic value of autophagy‐related genes (ARGs) and develop a ARG‐based model to evaluate 5‐year overall survival (OS) in BC patients. We acquired ARG expression profiling in a large BC cohort (N = 1007) from The Cancer Genome Atlas (TCGA) database. The correlation between ARGs and OS was confirmed by the LASSO and Cox regression analyses. A predictive model was established based on independent prognostic variables. Thus, time‐dependent receiver operating curve (ROC), calibration plot, decision curve and subgroup analysis were conducted to determine the predictive performance of ARG‐based model. Four ARGs (ATG4A, IFNG, NRG1 and SERPINA1) were identified using the LASSO and multivariate Cox regression analyses. A ARG‐based model was constructed based on the four ARGs and two clinicopathological risk factors (age and TNM stage), dividing patients into high‐risk and low‐risk groups. The 5‐year OS of patients in the low‐risk group was higher than that in the high‐risk group (P  < 0.0001). Time‐dependent ROC at 5 years indicated that the four ARG–based tool had better prognostic accuracy than TNM stage in the training cohort (AUC: 0.731 vs 0.640, P  < 0.01) and validation cohort (AUC: 0.804 vs 0.671, P  < 0.01). The mutation frequencies of the four ARGs (ATG4A, IFNG, NRG1 and SERPINA1) were 0.9%, 2.8%, 8% and 1.3%, respectively. We built and verified a novel four ARG–based nomogram, a credible approach to predict 5‐year OS in BC, which can assist oncologists in determining effective therapeutic strategies.

中文翻译：

---

乳腺癌自噬相关预后基因的综合分析。

越来越多的证据表明，自噬在乳腺癌 (BC) 的进展中起着至关重要的作用。因此，本研究的目的是研究自噬相关基因 (ARG) 的预后价值，并开发基于 ARG 的模型来评估 BC 患者的 5 年总生存率 (OS)。我们从癌症基因组图谱 (TCGA) 数据库中获得了一个大型 BC 队列 (N = 1007) 中的 ARG 表达谱。LASSO 和 Cox 回归分析证实了 ARG 和 OS 之间的相关性。基于独立的预后变量建立预测模型。因此，进行了时间相关的受试者工作曲线（ROC）、校准图、决策曲线和亚组分析，以确定基于 ARG 的模型的预测性能。四种 ARG（ATG4A、IFNG、NRG1 和 SERPINA1) 使用 LASSO 和多变量 Cox 回归分析确定。基于四个 ARG 和两个临床病理学危险因素（年龄和 TNM 分期）构建基于 ARG 的模型，将患者分为高危组和低危组。低危组患者的 5 年 OS 高于高危组（P  < 0.0001)。5 年的时间依赖性 ROC 表明，在训练队列（AUC：0.731 对 0.640，P  < 0.01）和验证队列（AUC：0.804 对 0.671，P  < 0.01 ）中，四种基于 ARG 的工具具有更好的预后准确性）。四种ARG（ATG4A、IFNG、NRG1和SERPINA1）的突变频率分别为0.9%、2.8%、8%和1.3%。我们构建并验证了一种新的基于 4 ARG 的列线图，这是一种预测 BC 5 年 OS 的可靠方法，可以帮助肿瘤学家确定有效的治疗策略。