An effective fuzzy kernel clustering analysis approach for gene expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Mining

Background:

• Fuzzy clustering is crucial for analyzing microarray and gene expression data.
• Determining optimal parameters like cluster number and centers remains a challenge for existing fuzzy clustering methods.
• Gene expression data analysis requires robust and stable clustering techniques.

Purpose of the Study:

• To propose a novel fuzzy kernel clustering analysis (FKCA) approach for gene expression data.
• To enhance the identification of the optimal cluster number and cluster centers.
• To improve the stability and accuracy of clustering results for gene expression datasets.

Main Methods:

• Introduced a Gaussian kernel function to improve the Spectrum Analysis Method (SAM), creating an Improved SAM (ISAM) for optimizing characteristic differences and estimating cluster numbers.
• Developed a Maximum Distance Method (MDM) by combining subtractive clustering with max-min distance mean to determine cluster centers.
• Integrated ISAM and MDM into the FKCA framework to create an effective improved FKCA algorithm.

Main Results:

• The proposed ISAM and MDM methods demonstrated superiority and stability in experimental comparisons on gene expression data.
• The improved FKCA algorithm, incorporating ISAM and MDM, proved feasible for cluster analysis.
• Experimental results showed higher clustering accuracy compared to other related clustering algorithms on public gene expression and UCI database datasets.

Conclusions:

• The developed ISAM and MDM provide effective solutions for parameter selection in fuzzy clustering of gene expression data.
• The improved FKCA algorithm offers a more stable and accurate approach to analyzing gene expression patterns.
• This enhanced FKCA method holds significant potential for advancing bioinformatics and computational biology research.