面向癌症亚型预测的多组学AI模型
Multi-omics AI model for cancer subtype prediction
曹云芳 1李东喜1
作者信息
- 1. 太原理工大学大数据学院,山西晋中 030600
- 折叠
摘要
针对癌症亚型预测中仅使用单组学数据信息有限的问题,提出一种基于稀疏自编码器和相似网络融合的多组学癌症分型预测模型(multi-omics sparse auto-encoder,MOSAE).利用稀疏自编码器提取患者特征向量,应用相似网络融合方法构建患者的相似度网络.基于患者特征向量和患者相似度网络利用残差图卷积网络构建预测模型.实验结果表明,在乳腺癌和卵巢癌数据上,所提模型识别亚型的准确率比现有方法分别提高了 2.74％和19.74％.在TCGA的肺鳞状细胞癌和头颈部癌症数据上验证了 MOSAE模型的优越性.
Abstract
Aiming at the information limited issue of using only single omics data in predicting cancer subtypes,a multi-omics cancer classification prediction model(multi-omics sparse auto-encoder,MOSAE)based on sparse autoencoder and similar net-work fusion was proposed.Sparse autoencoder was used to extract patient feature vectors,and similarity network was construc-ted using similarity network fusion method.The residual graph convolution network was used to construct the prediction model based on the patient feature vector and the patient similarity network.Experimental results show that the accuracy of the pro-posed model is 2.74％and 19.74％higher than that of the existing methods in the identification of subtypes of breast cancer and ovarian cancer.The superiority of the MOSAE model is also verified on the data of lung squamous cell carcinoma and head and neck cancer of TCGA.
关键词
稀疏自编码器/残差图卷积网络/相似网络融合/多组学数据/癌症亚型/多模态/特征提取Key words
sparse autoencoder/residual graph convolutional network/similar network fusion/multi-omics data/cancer sub-type/multi-modal/feature extraction引用本文复制引用
基金项目
山西省基础研究计划基金项目(20210302124168)
山西省回国留学人员科研基金项目(2022-074)
出版年
2024
NETL
NSTL
万方数据