Gene and sample selection using T-score with sample selection.

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• Gene selection from high-dimensional microarray data presents statistical challenges.
• Filter-based gene selection methods are popular for their simplicity and efficiency.
• Sample selection has been overlooked in filter-based gene selection due to small sample sizes.

Purpose of the Study:

• To extend the simultaneous sample and gene selection approach.
• To develop a gene selection method that balances T-score and Support Vector Machine (SVM) based algorithms.
• To improve the performance of gene selection methods in terms of classification, stability, and redundancy.

Main Methods:

• A backward elimination method was employed.
• A modified logistic regression loss function was used for sample selection at each iteration.
• Selected samples were used to compute T-scores for gene ranking.

Main Results:

• The proposed method demonstrated improved computational complexity and stability compared to SVM-based methods.
• Classification performance was maintained without compromise.
• The approach showed effectiveness on both simulated and real gene-expression datasets.

Conclusions:

• The developed method offers an effective compromise between T-score and SVM-based algorithms for gene selection.
• Simultaneous sample and gene selection enhances the stability and efficiency of gene expression data analysis.
• This approach provides a valuable tool for identifying relevant genes in high-dimensional datasets.