摘要
      骨髓瘤（MM）是一种恶性浆细胞疾病，由于其耐药性而无法治愈。自噬在多发性骨髓瘤（MM）的稳态、生存和耐药性中发挥着不可或缺的作用。因此，本研究的目的是识别MM患者中潜在的自噬相关基因（ARGs）。我们下载了MM患者的转录组数据（GSE136400），以及来自基因表达综合库（GEO）的相应临床数据；将患者按6:4的比例随机分为两组，训练数据集中有212个样本，测试数据集中有142个样本。进行了多变量和单变量Cox回归分析，以确定自噬相关基因。单变量Cox回归分析表明，26个ARG与总生存率（OS）有显著相关性。我们构建了一个基于六个ARG:EIF2AK2（ENSG00000055332）、KIF5B（ENSG00000170759）、MYC（ENSG0000136997）、NRG2（ENSG00000 158458）、PINK1（ENSG0 0000158828）和VEGFA（ENSG0000 112715）的自噬相关风险预后模型，使用LASSO Cox回归分析来预测风险结果，与低风险队列相比，高风险队列的OS持续时间显著缩短。因此，基于ARG的模型显著预测了MM患者的预后，并在内部测试集中进行了验证。发现差异表达的基因主要富集在与炎症和免疫调节相关的途径中。肿瘤细胞的免疫浸润导致高危患者形成强大的免疫抑制微环境。随后通过蛋白质-药物网络分析分析了ARGs的潜在治疗靶点。因此，通过使用临床模型对ARGs进行综合分析，建立了MM的预后模型；我们进一步揭示了多发性骨髓瘤的分子景观特征。
Abstract
Myeloma (MM) is a malignant plasma cell disorder, which is incurable owing to its drug resistance. Autophagy performs an integral function in homeostasis, survival, and drug resistance in multiple myeloma (MM). Therefore, the purpose of the present research was to identify potential autophagy-related genes (ARGs) in patients with MM. We downloaded the transcriptomic data (GSE136400) of patients with MM, as well as the corresponding clinical data from the Gene Expression Omnibus (GEO); the patients were classified at random into two groups in a ratio of 6: 4, with 212 samples in the training dataset and 142 samples in the test dataset. Both multivariate and univariate Cox regression analyses were performed to identify autophagy-related genes. The univariate Cox regression analysis demonstrated that 26 ARGs had a significant correlation with overall survival (OS). We constructed an autophagy-related risk prognostic model based on six ARGs: EIF2AK2 (ENSG00000055332), KIF5B (ENSG00000170759), MYC (ENSG00000136997), NRG2 (ENSG00000158458), PINK1 (ENSG00000158828), and VEGFA (ENSG00000112715) using LASSO-Cox regression analysis to predict risk outcomes, which revealed substantially shortened OS duration in the high-risk cohort in contrast with that in the low-risk cohort. Therefore, the ARG-based model significantly predicted the MM patients’ prognoses and was verified in an internal test set. Differentially expressed genes were found to be predominantly enriched in pathways associated with inflammation and immune regulation. Immune infiltration of tumor cells resulted in the formation of a strong immunosuppressive microenvironment in high-risk patients. The potential therapeutic targets of ARGs were subsequently analyzed via protein–drug network analysis. Therefore, a prognostic model for MM was established via a comprehensive analysis of ARGs, through using the clinical models; we have further revealed the molecular landscape features of multiple myeloma.

https://www.hindawi.com/journals/jir/2022/3922739/