Tumor clustering using nonnegative matrix factorization with gene selection.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Tumor clustering is crucial for cancer class discovery.
• Nonnegative matrix factorization (NMF) is a promising clustering technique.
• Improving NMF performance for cancer research remains an objective.

Purpose of the Study:

• To enhance tumor clustering by integrating gene selection and sparsity into NMF.
• To evaluate the effectiveness of NMF extensions for cancer class discovery.

Main Methods:

• Independent component analysis (ICA) for gene selection to reduce noise.
• Applying NMF, sparse NMF, and NMF with sparseness constraints for clustering.
• Experimental evaluation varying cluster numbers and gene subset sizes.

Main Results:

• The proposed scheme integrating gene selection and sparse NMF improved clustering results.
• Independent component analysis effectively reduced the impact of irrelevant genes.
• Experiments demonstrated the robustness of the enhanced NMF approach across datasets.

Conclusions:

• Gene selection and sparsity enforcement significantly boost NMF-based tumor clustering.
• The developed method offers a more accurate approach to cancer class discovery.
• This work provides a refined computational tool for cancer genomics research.