A hybrid gene selection method based on gene scoring strategy and improved particle swarm optimization.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Gene selection is crucial for microarray data classification.
• Particle swarm optimization (PSO) is effective but prone to local minima, risking suboptimal gene selection.
• Selecting predictive genes with low redundancy while retaining key genes remains a challenge.

Purpose of the Study:

• To develop a hybrid gene selection method addressing PSO limitations.
• To enhance the selection of predictive genes with reduced redundancy.
• To improve the accuracy of microarray data classification.

Main Methods:

• A gene scoring strategy using randomization and extreme learning machine (ELM) filters irrelevant genes.
• An improved PSO incorporating the Metropolis criterion from simulated annealing to prevent premature convergence.
• Swarm reinitialization when trapped in local minima.

Main Results:

• The proposed method effectively filters irrelevant genes.
• Improved PSO successfully selects functional gene subsets sensitive to sample classes.
• The method achieves low redundancy in selected genes.

Conclusions:

• The hybrid approach combines gene scoring and improved PSO for robust gene selection.
• Selected gene subsets enable high prediction accuracy with ELM and support vector machine classifiers.
• The method demonstrates efficiency and effectiveness for microarray data analysis.