基于人工神经网络技术联合表观基因组学在肝癌诊断中的应用

Application of artificial neural network and epigenomics in the diagnosis ofhepatocellular carcinoma

目的:通过构建人工神经网络联合表观基因组学对肝癌组织及正常组织进行分类,为肝癌的诊断提供研究方向。方法:从TCGA数据库下载包含379例肝癌样本及50例正常肝组织样本的DNA甲基化数据集,筛选出差异高甲基化的、位于启动子区的、与对应基因呈负相关的CpG位点并使用Logistic回归分析筛选出与肝癌有关的4个CpG位点,将筛选出的位点使用人工神经网络技术对肝癌组织样本及正常肝组织样本进行分类。计算人工神经网络模型对样本判别的准确性、敏感度、特异度。结果:人工神经网络模型对样本判别的准确性为94.7%、敏感度为96.8%、特异度为92.7%。结论:人工神经网络模型可以很好地区分肝癌组织与正常组织,对肝癌诊断具有较高的价值。

Objective:To use the construction of artificial neural networks combined with epigenomics to classify liver cancer tissues and normal tissues to provide research directions for the diagnosis of liver cancer.Methods:The DNA methylation data of 379 HCC samples and 50 samples of regular liver tissues from the cancer genome atlas(TCGA)database was downloaded.The CpG sites that were located within the promoter region,with high differences in methylation and negatively correlated with their corresponding genes were selected.Logistic regression analysis was used to filter out the 4 CpG sites related to hepatocellular carcinoma.The hepatocellular carcinoma samples and regular liver samples were classified with artificial neural network.The accuracy,sensitivity,and specificity of artificial neural network models for sample discrimination were calculated.Results:The artificial neuron network model's prediction accuracy was 94.7%,the sensitivity was 96.8%and the specificity was 92.7%.Conclusion:The artificial neural network models could classify hepatocellular carcinoma tissue and normal tissue.It is valuable to the diagnosis of hepatocellular carcinoma.