Multiobjective binary biogeography based optimization for feature selection using gene expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Healthcare

Background:

• Gene expression data are crucial for cancer diagnosis and classification but often contain noise and redundancy.
• Identifying a subset of highly discriminative genes is essential for accurate classification and efficient analysis.

Purpose of the Study:

• To propose a novel multi-objective biogeography-based optimization (MOBBBO) algorithm for selecting informative genes from gene expression data.
• To enhance the classification accuracy of cancer samples by identifying a small, relevant subset of genes.

Main Methods:

• Utilized Fisher-Markov selector to pre-select top 60 gene expression data.
• Developed Binary Biogeography-Based Optimization (BBBO) for discrete problems, incorporating binary migration and mutation models.
• Integrated non-dominated sorting and crowding distance methods into BBBO to create MOBBBO.
• Employed Support Vector Machine (SVM) with Leave-One-Out Cross-Validation (LOOCV) for classification.

Main Results:

• The MOBBBO method demonstrated effectiveness in selecting a small subset of informative genes.
• Experimental results on ten gene expression dataset benchmarks showed superior or comparable performance against existing Particle Swarm Optimization (PSO) and SVM methods.
• The proposed gene selection approach improved classification quality.

Conclusions:

• The MOBBBO algorithm is an effective and efficient method for gene selection in cancer classification.
• This approach offers a promising tool for bioinformatics, aiding in the development of better diagnostic and classification strategies for cancer.
• The study highlights the potential of multi-objective optimization techniques in handling complex biological data for improved clinical applications.