Gene selection for microarray gene expression classification using Bayesian Lasso quantile regression.

Area of Science:

• Bioinformatics
• Statistical Genetics
• Machine Learning

Background:

• Gene selection is crucial for improving classification accuracy in biological data analysis.
• Current gene selection methods struggle with data anomalies like outliers and heteroskedasticity.
• Robust gene selection is needed for reliable binary classification regression tasks.

Purpose of the Study:

• To propose a novel Bayesian hierarchical model for robust gene selection.
• To address limitations of existing methods sensitive to data anomalies.
• To enhance classification accuracy in gene expression data analysis.

Main Methods:

• Development of a new Bayesian Lasso method incorporating a skewed Laplace distribution for errors.
• Utilizing a scaled mixture of uniform distribution for regression parameters.
• Employing Bayesian Markov Chain Monte Carlo (MCMC) estimation for model fitting.

Main Results:

• The proposed Bayesian hierarchical model demonstrates superior performance compared to existing methods.
• Experimental results on four benchmark gene expression datasets confirm the method's effectiveness.
• The new Bayesian Lasso approach successfully identifies the most relevant genes.

Conclusions:

• The proposed Bayesian hierarchical model offers a robust and effective solution for gene selection.
• This method significantly improves classification accuracy by overcoming data anomaly issues.
• The approach provides a valuable tool for analyzing gene expression data in bioinformatics.