基于包含邻居信息的相似网络融合对癌症亚型进行聚类

发布时间：2018-05-09 19:45

  本文选题：聚类 + 网络融合　； 参考：《中国海洋大学学报(自然科学版)》2017年S1期

【摘要】：近年来随着人类基因图谱计划和癌症基因图谱计划的实施,大量的癌症数据集体涌现。如何将这些数据有效地集合起来,利用其互补性来区分癌症亚型变得尤为重要。现存在的很多方法大都根据单一的数据类型对癌症亚型进行聚类,这些方法忽略了数据之间交互影响的信息。相似网络融合(Similarity Network Fusion(SNF))是一种可以把不同数据类型融合到一起的有效方法,其中构建样本之间的相似网络是该方法的重要步骤之一。本文提出包含邻居信息的相似网络融合(Neighborhood-Information-Embedded Similarity Network Fusion(NSNF))方法,用包含邻居信息的多重紧密k近邻方法代替原有的k近邻方法来构建相似网络,并将其运用于癌症亚型聚类。最后用4种癌症的实验数据证明了提出的NSNF方法比传统的SNF方法在聚类性能上有了很大的提高。
[Abstract]:In recent years, with the implementation of the human gene mapping program and the cancer gene mapping program, a large number of cancer data have emerged. How to effectively aggregate these data and make use of their complementarities to distinguish cancer subtypes is particularly important. There are many existing methods to cluster cancer subtypes according to a single data type. These methods ignore the information of data interaction. Similarity Network fusion is an effective method to fuse different data types together, and the construction of similar networks between samples is one of the important steps of this method. In this paper, a similar network fusion method containing neighbor information is proposed. The method of neighbor information is used to construct the similar network instead of the original k-nearest neighbor method, and it is applied to cancer subtype clustering. Finally, the experimental data of four kinds of cancers show that the proposed NSNF method is better than the traditional SNF method in clustering performance.
【作者单位】： 中国海洋大学数学科学学院;青岛市市立医院;
【基金】：国家自然科学基金项目(11271341)资助~~
【分类号】：R73-3
，

本文编号：1867175