Hybrid gene selection approach using XGBoost and multi-objective genetic algorithm for cancer classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray gene expression datasets present challenges due to high dimensionality (many genes) and low sample size.
• Identifying cancer-relevant genes from large datasets is crucial but difficult, impacting classification accuracy.

Purpose of the Study:

• To propose a novel two-stage gene selection approach, XGBoost-MOGA, for enhanced cancer classification in microarray data.
• To improve the accuracy and efficiency of identifying biologically significant genes for cancer subtyping.

Main Methods:

• The first stage employs extreme gradient boosting (XGBoost) for ensemble-based feature selection to rank and filter genes.
• The second stage utilizes a multi-objective optimization genetic algorithm (XGBoost-MOGA) to find an optimal subset from the most relevant genes.

Main Results:

• XGBoost-MOGA demonstrated superior performance compared to existing state-of-the-art feature selection methods.
• The proposed method achieved significant improvements in accuracy, F-score, precision, and recall across 14 microarray datasets.

Conclusions:

• The XGBoost-MOGA approach offers a robust and effective solution for gene selection in cancer classification.
• This method enhances the identification of key genes, leading to more accurate and reliable cancer diagnosis from gene expression data.