Classification by integrating plant stress response gene expression data with biological knowledge.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Microarray data classification is challenging due to the high dimensionality of gene numbers.
• Existing clustering methods for gene expression data often lack biological interpretability.
• Integrating biological knowledge can improve the biological relevance of gene clusters.

Purpose of the Study:

• To develop and evaluate a novel clustering method for plant stress response gene expression data.
• To integrate biological knowledge, specifically Gene Ontology (GO) semantic similarity, with gene expression data.
• To enhance the biological interpretability and predictive performance of genomic analysis.

Main Methods:

• Affinity propagation clustering algorithm was employed to group genes based on expression data and GO semantic similarity.
• Neighborhood rough set was utilized for the selection of representative genes from identified clusters.
• Classification models were built using reduced gene subsets to assess the effectiveness of the proposed method.

Main Results:

• The integrated approach successfully generated biologically meaningful gene clusters.
• The method demonstrated superior performance in prediction accuracy compared to classical methods when using reduced gene subsets.
• Quantitative analysis confirmed the effectiveness of information fusion in selecting biologically significant gene subsets.

Conclusions:

• Integrating biological knowledge with gene expression data significantly improves the interpretability of clustering results.
• The proposed method offers an effective strategy for attribute reduction in high-dimensional genomic data.
• This approach facilitates the selection of biologically significant genes for further research and application.