Gene selection for tumor classification using neighborhood rough sets and entropy measures.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Tumor classification using gene expression data is crucial for cancer diagnosis.
• Gene selection is a key challenge due to high dimensionality and limited samples in gene expression datasets.
• Traditional rough set methods struggle with real-value and noisy data, often requiring discretization that reduces accuracy.

Purpose of the Study:

• To propose a novel gene selection method for tumor classification using gene expression data.
• To address the limitations of traditional rough set methods in handling real-value and noisy data.
• To improve the accuracy of tumor classification by discovering compact and informative gene subsets.

Main Methods:

• Developed a neighborhood rough set model capable of processing real-value gene expression data without discretization.
• Introduced an entropy measure within the neighborhood rough set framework to manage data uncertainty and noise.
• Designed a gene selection algorithm integrating neighborhood granules and the proposed entropy measure.

Main Results:

• The proposed neighborhood rough set model effectively handles real-value and noisy gene expression data.
• The entropy measure aids in identifying compact gene subsets by addressing data uncertainty.
• Experimental results on two gene expression datasets demonstrate improved tumor classification accuracy.

Conclusions:

• The novel gene selection method based on neighborhood rough sets and entropy measure is effective for tumor classification.
• This approach enhances accuracy by preserving gene classification information and managing data noise.
• The method offers a robust alternative for gene selection in complex biological datasets.