Co-differential Gene Selection and Clustering Based on Graph Regularized Multi-View NMF in Cancer Genomic Data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer genomic data offers complementary insights from diverse sources.
• Feature selection and multi-view clustering are crucial for leveraging this data.
• Existing methods may not fully exploit the geometric structure within multi-view data.

Purpose of the Study:

• To propose a novel integrated model for selecting common differential genes and performing multi-view clustering.
• To enhance the model by incorporating geometric information from multi-view genomic data.
• To improve the understanding of cancer genetic activity through advanced bioinformatics techniques.

Main Methods:

• Development of Multi-view Non-negative Matrix Factorization (MvNMF).
• Introduction of Graph Regularized MvNMF (GMvNMF) by adding graph regularization.
• Application of GMvNMF to four multi-view genomic datasets.

Main Results:

• GMvNMF effectively identifies shared feature and cluster structures.
• The method captures the manifold structure of multi-view genomic data.
• GMvNMF demonstrates superior performance compared to existing representative methods in experiments.

Conclusions:

• GMvNMF is a powerful tool for analyzing multi-view cancer genomic data.
• The integration of graph regularization enhances the model's ability to capture data structure.
• This approach offers a promising direction for cancer research and biomarker discovery.