Feature selection for high dimensional microarray gene expression data via weighted signal to noise ratio.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• High-dimensional gene expression data presents challenges in dimensionality reduction and computational efficiency for classification.
• Effective feature selection is crucial for identifying biologically relevant genes and improving model performance.

Purpose of the Study:

• To introduce a novel feature selection method, weighted signal to noise ratio (WSNR), for identifying informative genes in high-dimensional classification problems.
• To evaluate the performance of WSNR against existing feature selection techniques using multiple gene expression datasets.

Main Methods:

• The weighted signal to noise ratio (WSNR) method combines weights derived from support vectors and signal-to-noise ratio.
• Feature weights are multiplied and ranked in descending order, with higher weights indicating greater discriminatory power.
• The method was validated on eight diverse gene expression datasets and simulated data.

Main Results:

• The WSNR method outperformed four established feature selection methods on 6 out of 8 real-world gene expression datasets.
• Comparative analysis using box-plots and bar-plots visually demonstrated WSNR's superior performance.
• Simulation analysis confirmed that WSNR achieved better results than all compared methods.

Conclusions:

• The proposed weighted signal to noise ratio (WSNR) method is a highly effective approach for feature selection in high-dimensional gene expression data.
• WSNR offers improved accuracy and efficiency in identifying key genes for classification tasks.
• The method shows significant potential for applications in bioinformatics and computational biology research.