机器学习识别肝细胞癌骨转移驱动基因及体外放疗分析  

Recognition of bone metastasis driving genes in hepatocellular carcinoma by machine learning and analysis of in external-beam radiotherapy

作　　者：王瑜[1] 童列杏 康婉英 丁鑫 钱和生[4] Wang Yu;Tong Liexing;Kang Wanying;Ding Xin;Qian Hesheng(Department of Radiotherapy,Fuyang Cancer Hospital,Fuyang 236000,Anhui,China;Department of Laboratory,Fuyang Cancer Hospital,Fuyang 236000,Anhui,China;Department of Imaging,Fuyang Cancer Hospital,Fuyang 236000,Anhui,China;Department of Oncology,Fuyang Cancer Hospital,Fuyang Cancer Hospital,Fuyang 236000,Anhui,China)

机构地区：[1]阜阳市肿瘤医院放疗科,安徽阜阳236000 [2]阜阳市肿瘤医院检验科,安徽阜阳236000 [3]阜阳市肿瘤医院影像科,安徽阜阳236000 [4]阜阳市肿瘤医院肿瘤内科,安徽阜阳236000

出　　处：《肝癌电子杂志》2023年第3期64-75,共12页Electronic Journal of Liver Tumor

摘　　要：目的:通过机器学习识别肝细胞癌骨转移(HCC-BM)驱动基因并初步探索体外放疗(EBRT)对驱动基因表达水平影响。方法:基因表达综合数据库中GSE14520和GSE84402数据集合并作为测试集,GSE121248和GSE54236数据集分别作为验证集。执行最小绝对收缩和选择操作符,以及支持向量机-回归特征消除筛选关键基因,获得与GeneCards数据库中HCC-BM相关基因有交集的HCC-BM驱动基因。使用CIBERSORT评估HCC-BM驱动基因与免疫细胞浸润间相关性。选取阜阳市肿瘤医院2020—2021年间收治的22例HCC-BM患者为研究组,同期选取我院22例正常体检的健康人群作为对照组。比较EBRT前后HCC-BM驱动基因表达水平变化,以及ASPM表达水平与疼痛缓解关系。结果:4个HCC-BM驱动基因被识别,分别为异常纺锤体样小头畸形相关蛋白(ASPM)、C型凝集素结构域家族4成员(CLEC4M)、细胞外基质蛋白1(ECM1)、无花果酶2(FCN2)。测试集中相邻肝正常组织ASPM表达较HCC组织降低(P<0.05),CLEC4M、ECM1和FCN2表达较HCC组织增高(P<0.05);验证集均获得同样结果。ASPM表达较高HCC患者预后较差,而CLEC4M、ECM1和FCN2表达较低患者预后较好。ASPM与M2型巨噬细胞呈正相关;CLEC4与M2型巨噬细胞、中性粒细胞和M1型巨噬细胞呈正相关;FCN2与M0型巨噬细胞呈负相关。研究组血清ASPM表达高于对照组(P<0.05),FCN2表达低于对照组(P<0.05)。EBRT后HCC-BM患者血清ASPM表达水平降低(P<0.05)。完全缓解者血清ASPM表达水平低于部分缓解者。结论:ASPM、CLEC4M、ECM1、FCN2可能参与HCC-BM发生机制,其中ASPM可能与EBRT效果相关。进一步探索驱动基因有助于开发治疗靶点,提高EBRT效果,改善预后。Objective:To identify the driving genes of hepatocellular carcinoma with bone metastasis(HCC-BM)by machine learning and to explore the effect of external-beam radiotherapy(EBRT)on the expression level of driving genes.Methods:GSE14520 and GSE84402 data sets in Gene Expression Omnibus database were combined as test sets,and GSE121248 andGSE54236 data sets were used as validation sets respectively.The least absolute shrinkage and selection operator and support vector machine-regression feature elimination were performed to screen the key differentially expressed genes,and the intersection with HCC-BM related genes in the GeneCards database obtained HCC-BM driving genes.CIBERSORT evaluated the correlation between HCC-BM driver gene and immune cell infiltration.The 22 HCC-BM patients admitted to our hospital from 2020 to 2021 were selected as the study group.In the same period,22 healthy people with normal physical examinations in our hospital were selected as the control group.The changes in HCC-BM driver gene expression before and after EBRT were compared,and the relationship between HCC-BM driver gene expression and and relationship with ASPM expression levels and pain relief.Results:Four HCC-BM driving genes were identified,it is respectively assembly factor for spindle microtubules(ASPM),C-type lectin domain family 4 member M(CLEC4M),extracellular matrix protein 1(ECM1),and ficolin 2(FCN2).The expression of ASPM in normal liver tissue was lower than that in HCC tissue(P<0.05),and the expression of clec4m,ECM1,and FCN2 was higher than that in HCC tissue(P<0.05).The same results were obtained in both validation sets GSE121248 and GSE54236.The prognosis of HCC patients with high ASPM expression is poor,while patients with low CLEC4M,ECM1 and FCN2 expression are better.ASPM was positively correlated with macrophages m2;CLEC4M was positively correlated with macrophages M2,neutrophils,and M1 macrophages;FCN2 was negatively correlated with macrophages M0.The expression of serum ASPM in 22 patients with HCC-BM was highe

关 键 词：肝细胞癌 骨转移 免疫细胞 体外放射治疗 预后 

分 类 号：R735.7[医药卫生—肿瘤] R730.55[医药卫生—临床医学]