Hybrid Gene Selection Method Using Graph Theory and Chaotic Bee Colony Optimization

Abstract

Gene expression data in bioinformatics often suffer from high dimensionality and limited size, impacting the efficacy of data mining and machine learning algorithms. Gene selection methods aim to mitigate this issue by identifying relevant genes while discarding irrelevant or redundant ones. Traditional methods may struggle with accuracy and efficiency in selecting optimal gene subsets. This paper introduces a hybrid approach combining graph theory and Chaotic Bee Colony Optimization (CBCO) for gene selection. Initially, a filter method based on Fisher score reduces the gene pool. Next, genes are represented as nodes in a graph, where relationships construct edges. Graph K-means clustering groups genes into clusters, enhancing diversity. The CBCO algorithm then optimizes gene subset selection based on multiple criteria: classification error, node and edge centrality, specificity, and number of genes selected. A repair operator ensures at least one gene per cluster is chosen, enhancing overall solution robustness. Evaluation on datasets shows a superior classification accuracy and reduced gene selection compared to state-of-the-art methods. For instance, the proposed method achieves an average accuracy improvement of 5% and reduces gene selection by 30% across datasets. The hybrid method effectively addresses gene selection challenges by integrating graph-based clustering and multi-objective CBCO optimization. It surpasses existing techniques by enhancing classification accuracy and reducing computational overhead, demonstrating its potential for improving bioinformatics analyses.