Combining Pareto-optimal clusters using supervised learning for identifying co-expressed genes

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Microarrays enable simultaneous analysis of gene transcription levels, driving the need for effective gene clustering.
• Clustering algorithms are crucial for identifying co-expressed genes in large-scale biological datasets.

Purpose of the Study:

• To address fuzzy clustering in microarray data as a multi-objective optimization problem.
• To develop a novel fuzzy majority voting approach for combining Pareto-optimal clustering solutions.
• To enhance the identification of biologically relevant co-expressed gene clusters.

Main Methods:

• Formulating fuzzy clustering as a multi-objective optimization problem to find Pareto-optimal solutions.
• Implementing a fuzzy majority voting strategy to integrate information from multiple clustering outcomes.
• Utilizing a Support Vector Machine (SVM) classifier for supervised learning on identified gene sets.

Main Results:

• Demonstrated performance on five benchmark microarray datasets (Yeast, Arabidopsis, Human, Rat).
• Comparative analysis with different SVM kernels and existing clustering techniques.
• Statistical and biological significance tests confirmed the superiority and relevance of the proposed approach.

Conclusions:

• The proposed method efficiently identifies co-expressed genes in microarray data.
• Generated gene clusters are biologically significant, grouping genes with similar functions.
• The approach offers an effective tool for analyzing gene expression patterns.