A Multi-Task Ensemble Strategy for Gene Selection and Cancer Classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene expression profiling is crucial for tumor classification.
• High dimensionality and limited samples in gene expression data pose challenges for accurate tumor classification.
• Existing gene selection methods often lack stability and consistency, especially with small datasets.

Purpose of the Study:

• To develop a stable and consistent gene selection method for improved tumor classification using gene expression data.
• To address the limitations of existing methods in handling high-dimensional and small-sample datasets.
• To enhance both classification performance and model interpretability.

Main Methods:

• A multi-task ensemble strategy combining repeated sampling with joint feature selection and classification.
• Application of multi-task logistic regression with ℓ2,1 group sparsity regularization for consistent gene subset selection across tasks.
• Integration with standard classifiers like logistic regression and support vector machines within a single process.

Main Results:

• The proposed method demonstrates superior classification accuracy compared to baseline approaches.
• It achieves greater consistency in selected genes, indicating improved stability.
• Effective performance on both simulated and real-world gene expression datasets.

Conclusions:

• The multi-task ensemble strategy offers a robust solution for gene selection in tumor classification.
• This approach enhances the reliability and interpretability of gene expression-based tumor classification models.
• The method provides a significant advancement for analyzing complex genomic data in cancer research.