A Novel Gene Selection Method Based on Sparse Representation and Max-Relevance and Min-Redundancy.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene selection is crucial for data preprocessing in high-dimensional biological data.
• Maximum Relevance and Minimum Redundancy (MRMR) is a common gene selection method but overlooks gene interdependencies.
• Existing methods may not fully capture complex gene relationships for accurate classification.

Purpose of the Study:

• To propose a novel gene selection method, Sparse Representation and MRMR (SRCMRM), that addresses limitations of existing approaches.
• To enhance gene relevance and correlation evaluation by incorporating sparse representation.
• To improve classification accuracy in biological datasets through optimized gene selection.

Main Methods:

• The SRCMRM method combines sparse representation with the MRMR algorithm.
• It utilizes sparse representation coefficients to assess gene relevance to classification targets.
• A two-step process involves removing irrelevant genes and calculating gene-category correlations using sparse representation.

Main Results:

• The SRCMRM method demonstrated superior performance compared to various existing algorithms.
• Improved classification accuracy was consistently achieved across all tested datasets.
• Experimental validation confirmed the effectiveness of the proposed SRCMRM approach.

Conclusions:

• The SRCMRM method offers a robust and effective solution for gene selection.
• Its ability to capture gene interdependencies leads to enhanced classification accuracy.
• The method holds practical significance for bioinformatics and computational biology applications.