M2型巨噬细胞特征基因与结肠癌免疫微环境研究

M2-type macrophage signature genes and the immune microenvironment of colon cancer

目的通过机器学习结合生物信息学技术以筛选核心预后价值的巨噬细胞的特征基因,并且探索其与免疫微环境、肿瘤免疫治疗的关系.方法本研究搜集TCGA数据库中COAD数据集和GEO数据库中(GSE39582)数据集.CIBERSORT法计算肿瘤样本中M2型巨噬细胞水平,通过相关性分析、单因素和多因素Cox回归分析和随机生存森林算法筛选特征基因.ESTIMATE算法计算肿瘤样本的免疫微环境评分(间质评分与免疫评分),并且研究特征基因及其关系,最后在免疫治疗队列中进行验证.结果本研究确定PPM1M与MRAS作为机器学习确定的核心预后基因.在TCGA数据中,高表达水平MRAS人群拥有更短的无进展生存时间(P=0.0013).在GEO数据中,高表达PPM1M基因(P=0.031)和MRAS基因(P=0.002)均与复发相关.PPM1M、MRAS基因和肿瘤免疫评分和间质评分均呈正相关、与抑制性调节T淋巴细胞水平呈正相关.最后,在免疫治疗评价中,高表达的PPM1M和MRAS患者接受免疫治疗的预后更好.结论机器学习确定的M2型巨噬细胞特征基因与生存、复发和进展相关.在免疫微环境中,PPM1M和MRAS均与抑制性的肿瘤免疫成分和间质成分呈正相关.此外,PPM1M和MRAS可能成为免疫治疗疗效的新型标志物.

Objective We aimed to identify for macrophage-2(M2)characteristic genes with hub prognostic value through machine learning combined with bioinformatics techniques,and to explore their relationship with the immune microenvironment and tumor immunotherapy.Methods This study collected the TCGA-COAD dataset and the dataset(GSE39582)from the GEO database.The CIBERSORT method was used to calculate the levels of M2-type macrophages in tumor samples,and characteristic genes were screened through correlation analysis,univariate and multivariate Cox regression analysis,and the random survival forest algorithm.The ESTIMATE algorithm was employed to calculate the immune microenvironment scores(stromal score and immune score)of tumor samples,and to study the characteristic genes and their relationships,finally validating in an immunotherapy cohort.Results This study identified PPM1M and MRAS as core prognostic genes determined by machine learning.In the TCGA data,populations with high expression levels of MRAS had shorter progression-free survival(P=0.0013).In the GEO data,high expression of PPM1M gene(P=0.031)and MRAS gene(P=0.002)were both associated with recurrence.Both PPM1M and MRAS genes were positively correlated with tumor immune score and stromal score,and positively correlated with the levels of suppressive regulatory T cells(Treg).Finally,in the evaluation of immunotherapy,patients with high expression of PPM1M and MRAS had better prognosis after receiving immunotherapy.Conclusion Characteristic genes of M2-type macrophages determined by machine learning are related to survival,recurrence,and progression.In the immune microenvironment,PPM1M and MRAS are both positively correlated with suppressive tumor immune components and stromal components.Furthermore,PPM1M and MRAS may serve as novel biomarkers for the efficacy of immunotherapy.