%0 Journal Article
%T 基于深度多视图对比学习方法的多组学数据整合及预后预测模型构建
Integration of Multi-Omics Data and Prognostic Prediction Model Construction Based on Deep Multi-View Contrastive Learning Methods
%A 高新凤
%J Advances in Applied Mathematics
%P 4182-4190
%@ 2324-8009
%D 2024
%I Hans Publishing
%R 10.12677/aam.2024.139399
%X 在癌症研究中，精准识别癌症亚型和评估患者预后对制定优化治疗方案至关重要。高通量测序技术生成的大量多组学数据为癌症预后研究提供了宝贵资源。深度学习方法能够有效整合这些数据，精确识别更多癌症亚型。在本研究中，我们分析了12种癌症的多组学数据集，并将其作为模型的输入。我们提出了一种基于卷积自动编码器的深度多视图对比学习模型(dmCLCAE)，该模型旨在利用多组学数据预测与生存相关的癌症亚型。为了验证模型的效果，我们对比了多组学因子分析算法(MOFA+)和深度学习模型(ProgCAE)在不同癌症类型分类中的表现。结果显示，dmCLCAE在区分不同生存亚型方面表现出显著优势，同时在预测一致性上也有更优异的表现。
In cancer research, accurately identifying cancer subtypes and assessing patient prognosis are crucial for developing optimized treatment strategies. The vast amount of multi-omics data generated by high-throughput sequencing technologies provides valuable resources for cancer prognosis studies. Deep learning methods can effectively integrate these data to accurately identify more cancer subtypes. In this study, we analyzed multi-omics datasets from 12 types of cancer and used them as input for our model. We proposed a deep multi-view contrastive learning model based on a convolutional autoencoder (dmCLCAE), designed to predict survival-related cancer subtypes using multi-omics data. To validate the model’s performance, we compared it with the Multi-Omics Factor Analysis v2 (MOFA+) and prognostic model based on a convolutional autoencoder (ProgCAE) in classifying various cancer types. The results showed that dmCLCAE demonstrated a significant advantage in distinguishing different survival subtypes and exhibited superior consistency in predictions.
%K 多组学数据，
%K 卷积自编码器，
%K 对比学习，
%K 深度学习
Multi-Omics Data
%K Convolutional Autoencoder
%K Contrastive Learning
%K Deep Learning
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=96179