Simulated annealing aided genetic algorithm for gene selection from microarray data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray gene expression datasets are crucial for cancer biomarker discovery.
• High dimensionality and gene-to-sample ratios in these datasets necessitate effective gene filtering.
• Redundant data can obscure the identification of truly informative genes.

Purpose of the Study:

• To propose a novel meta-heuristic algorithm, Simulated Annealing aided Genetic Algorithm (SAGA), for identifying informative genes from high-dimensional datasets.
• To address the limitations of traditional Genetic Algorithms (GA), such as premature convergence and dependence on initial populations.
• To enhance gene selection accuracy for cancer biomarker identification.

Main Methods:

• Developed SAGA, integrating Simulated Annealing (SA) and Genetic Algorithm (GA) for a balanced search.
• Employed clustering-based population generation with SA to optimize GA's initial population distribution.
• Utilized the Mutually Informed Correlation Coefficient (MICC) for initial search space reduction.
• Evaluated SAGA on 6 microarray and 6 omics datasets.

Main Results:

• SAGA demonstrated superior performance compared to contemporary algorithms in gene identification.
• The algorithm effectively balances exploration and exploitation in the search space.
• Clustering-based population generation and MICC filtering enhanced SAGA's efficiency.

Conclusions:

• SAGA is a highly effective approach for identifying informative genes from high-dimensional omics data.
• The proposed method offers significant improvements over existing gene selection techniques.
• SAGA provides a robust tool for biomarker discovery in cancer research.