Journal
INTERDISCIPLINARY SCIENCES-COMPUTATIONAL LIFE SCIENCES
Volume 14, Issue 1, Pages 22-33Publisher
SPRINGER HEIDELBERG
DOI: 10.1007/s12539-021-00441-8
Keywords
Low-rank tensor; Hypergraph; Sample clustering; Tensor robust principal component analysis
Categories
Funding
- National Natural Science Foundation of China [61872220]
Ask authors/readers for more resources
Clustering analysis of cancer genomics data has been a focus in recent years. A new method, HTRPCA, using tensor and hypergraph to represent and save geometric structure information, shows effectiveness in clustering samples and outperforms other advanced methods, as demonstrated in experiments on TCGA datasets.
In recent years, clustering analysis of cancer genomics data has gained widespread attention. However, limited by the dimensions of the matrix, the traditional methods cannot fully mine the underlying geometric structure information in the data. Besides, noise and outliers inevitably exist in the data. To solve the above two problems, we come up with a new method which uses tensor to represent cancer omics data and applies hypergraph to save the geometric structure information in original data. This model is called hypergraph regularized tensor robust principal component analysis (HTRPCA). The data processed by HTRPCA becomes two parts, one of which is a low-rank component that contains pure underlying structure information between samples, and the other is some sparse interference points. So we can use the low-rank component for clustering. This model can retain complex geometric information between more sample points due to the addition of the hypergraph regularization. Through clustering, we can demonstrate the effectiveness of HTRPCA, and the experimental results on TCGA datasets demonstrate that HTRPCA precedes other advanced methods. [GRAPHICS]
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available