Towards Unsupervised Gene Selection: A Matrix Factorization Framework.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray gene expression analysis enables disease phenotype classification.
• Gene expression datasets contain numerous genes, many redundant or irrelevant for disease identification.
• Efficient gene selection is crucial for accurate disease classification.

Purpose of the Study:

• To propose a novel unsupervised two-stage method for coarse-fine gene selection.
• To address the challenge of redundant and insignificant genes in expression data analysis.
• To improve the efficiency and effectiveness of gene selection for disease classification.

Main Methods:

• An unsupervised two-stage approach combining K-means clustering and matrix factorization.
• Stage 1: K-means clustering to over-cluster genes and remove redundancy.
• Stage 2: Matrix factorization to select the most representative genes from the remaining set.

Main Results:

• The proposed method effectively selects significant genes from gene expression data.
• Experimental results on multiple datasets demonstrate the method's high performance.
• The approach successfully reduces data dimensionality while retaining crucial disease-related information.

Conclusions:

• The novel two-stage gene selection method is effective for disease phenotype classification.
• This approach offers an efficient way to handle large, complex gene expression datasets.
• The technique provides a valuable tool for advancing gene expression data analysis in disease research.