A Review of Matched-pairs Feature Selection Methods for Gene Expression Data Analysis

Area of Science:

• Bioinformatics and computational biology
• Genomics and transcriptomics
• Biostatistics

Background:

• High-dimensional gene expression data from technologies like RNA-sequencing requires effective feature selection for analysis.
• Traditional feature selection methods face challenges, especially with paired gene expression data from matched case-control designs (MCCD).
• Matched-pairs feature selection (MPFS) methods are underdeveloped but offer potential for increased efficiency in analyzing paired data.

Purpose of the Study:

• To review and compare the performance of existing matched-pairs feature selection (MPFS) methods.
• To analyze the algorithmic complexity of these MPFS methods.
• To provide guidance for researchers in selecting appropriate MPFS methods for their gene expression data analyses.

Main Methods:

• Comparison of 10 feature selection methods, including eight MPFS methods and two traditional unpaired methods.
• Application of these methods to two real-world gene expression datasets.
• Evaluation of method performance using three distinct classification algorithms.
• Analysis of algorithm complexity through program execution.

Main Results:

• Performance comparison of eight MPFS methods and two unpaired methods on real datasets.
• Assessment of computational complexity for each feature selection algorithm.
• Identification of characteristics and potential advantages of MPFS methods.

Conclusions:

• MPFS methods show promise for analyzing paired gene expression data.
• Understanding method performance and complexity is crucial for effective feature selection.
• This review provides a foundation for choosing appropriate MPFS techniques in bioinformatics.