Gene selection based on adaptive neighborhood-preserving multi-objective particle swarm optimization.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• High-dimensional gene expression data analysis faces challenges like dimensionality and computational burden.
• Traditional methods struggle with inconsistent results and preserving local data structures, limiting accuracy and interpretability.

Purpose of the Study:

• To propose an adaptive neighborhood-preserving multi-objective particle swarm optimization (ANPMOPSO) framework for effective gene selection.
• To address limitations in dimensionality reduction, population diversity, exploration-exploitation balance, and biological relevance prioritization.

Main Methods:

• Developed ANPMOPSO with weighted neighborhood-preserving ensemble embedding (WNPEE) for dimensionality reduction.
• Incorporated Sobol sequence (SS) initialization for enhanced diversity and differential evolution (DE) for adaptive velocity updates.
• Introduced a novel ranking strategy combining Pareto dominance and neighborhood preservation quality.

Main Results:

• ANPMOPSO achieved 100% classification accuracy on Leukemia and SRBCT datasets using only 3-5 genes, outperforming competitors by 5-15%.
• Reduced gene subsets by 40-60% while improving accuracy.
• Demonstrated superior performance on multi-modal test functions, with higher hypervolume values indicating better convergence and diversity.

Conclusions:

• ANPMOPSO offers a robust solution for high-dimensional gene selection, balancing computational cost and biological relevance.
• The framework shows significant improvements in predictive accuracy and biological interpretability compared to state-of-the-art methods.
• ANPMOPSO is a promising tool for advancing gene selection in bioinformatics research.