Feature selection via robust weighted score for high dimensional binary class-imbalanced gene expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Genomics

Background:

• High-dimensional gene expression data often presents class-imbalance challenges.
• Skewed class distributions negatively impact the performance of classification algorithms.
• Effective feature selection is crucial for accurate binary classification in genomics.

Purpose of the Study:

• To propose a robust weighted score for unbalanced data (ROWSU) for feature selection.
• To address the challenge of class imbalance in high-dimensional gene expression datasets.
• To improve the performance of classification algorithms on imbalanced genomic data.

Main Methods:

• Data balancing through synthetic minority over-sampling technique.
• Greedy search approach for initial minimum gene subset selection.
• Novel weighted robust score using support vector weights for gene refinement.

Main Results:

• The ROWSU method successfully selects discriminative genes from imbalanced datasets.
• Evaluated on 7 gene expression datasets, ROWSU demonstrated superior performance.
• Outperformed existing methods in classification accuracy, sensitivity, and F1-score using kNN and RF classifiers.

Conclusions:

• The proposed ROWSU method is effective for feature selection in imbalanced gene expression data.
• ROWSU enhances classifier performance by selecting the most discriminative genes.
• This approach offers a robust solution for binary classification problems in genomics with skewed data.