A Gene Selection Method for Microarray Data Based on Binary PSO Encoding Gene-to-Class Sensitivity Information.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Genomics

Background:

• Traditional gene selection methods for microarray data often yield poorly interpretable results.
• Existing approaches primarily focus on predictive accuracy or data distribution, neglecting biological relevance.

Purpose of the Study:

• To develop an improved gene selection method that enhances both prediction accuracy and the interpretability of selected genes.
• To integrate prior biological information into the gene selection process for more meaningful results.

Main Methods:

• A novel gene selection approach combining binary particle swarm optimization (BPSO) with gene-to-class sensitivity (GCS) information.
• Gene-to-class sensitivity (GCS) is extracted using extreme learning machine (ELM) and encoded into BPSO for initializing particles, updating, velocity modification, and adaptive mutation.
• The method selects functional gene subsets highly sensitive to sample classes.

Main Results:

• The proposed method successfully identified a small subset of discriminative genes.
• Classifiers including ELM, K-nearest neighbor, and support vector machine achieved high prediction accuracy using the selected genes.
• The effectiveness and efficiency of the gene selection method were validated on five public microarray datasets.

Conclusions:

• The proposed BPSO-based gene selection method effectively identifies functionally relevant and discriminative genes from microarray data.
• This approach significantly improves prediction accuracy while ensuring better interpretability of the selected genes.
• The integration of GCS information enhances the biological significance of gene selection in bioinformatics.