Comparison of data-merging methods with SVM attribute selection and classification in breast cancer gene expression

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• DNA microarrays identify potential prognostic gene markers.
• Limited sample sizes in individual studies lead to unstable gene signatures and reduced performance.
• Integrating multiple studies increases sample size and aims to improve signature reliability.

Purpose of the Study:

• To investigate the impact of merging DNA microarray datasets on breast cancer classification.
• To evaluate if data integration enhances the performance of Support Vector Machine (SVM) based prognostic models.
• To assess the stability and predictive accuracy of gene signatures derived from integrated datasets.

Main Methods:

• Utilized three distinct breast cancer DNA microarray datasets.
• Employed Support Vector Machine (SVM) with attribute selection for classification.
• Focused on predicting distant metastasis using integrated and individual datasets.

Main Results:

• Breast cancer classification performance did not show improvement after merging datasets.
• Gene signatures derived from integrated data were not more stable or predictive than those from individual studies.
• The findings align with previous research using different methodologies.

Conclusions:

• Data merging does not benefit breast cancer classification based on DNA microarray studies.
• The instability and performance limitations of gene signatures persist even with increased sample size through integration.
• Further research may be needed to explore alternative methods for improving prognostic marker identification.