Two-stage feature selection for classification of gene expression data based on an improved Salp Swarm Algorithm.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray technology generates ultra-high dimensional gene expression data, posing challenges for identifying key genes.
• High-dimensional biomedical data, especially with small sample sizes, requires robust feature selection to avoid the curse of dimensionality.

Purpose of the Study:

• To propose a novel two-stage feature selection framework for high-dimensional gene expression data.
• To develop an improved gene selection method using an enhanced binary Salp Swarm Algorithm.
• To evaluate the framework's performance in terms of gene selection and classification accuracy.

Main Methods:

• A two-stage framework combining filtering, embedding, and wrapper methods.
• Stage one utilizes Weighted Gene Co-expression Network Analysis (WGCNA), Random Forest, and mRMR.
• Stage two employs an improved binary Salp Swarm Algorithm with machine learning for adaptive feature subset selection.

Main Results:

• The proposed framework achieved classification accuracy equal to or higher than other advanced intelligent algorithms across 10 datasets.
• Achieved over 97.6% classification accuracy on all tested datasets.
• Demonstrated effectiveness in solving feature selection problems for high-dimensional data.

Conclusions:

• The proposed two-stage feature selection framework effectively handles high-dimensional gene expression data.
• The novel gene selection method based on the improved binary Salp Swarm Algorithm is highly effective.
• The framework shows no dataset limitations and has broad applicability in feature selection across various fields.