An NMF-L2,1-Norm Constraint Method for Characteristic Gene Selection.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene expression data present challenges like high dimensionality, noise, and non-sparsity.
• Existing methods address only some of these data complexities.
• Effective characteristic gene selection is crucial for biological insights.

Purpose of the Study:

• To develop a unified method for characteristic gene selection that addresses multiple challenges in gene expression data.
• To improve the robustness and sparsity of gene selection methods.
• To enhance the identification of characteristic genes compared to existing approaches.

Main Methods:

• Nonnegative matrix factorization via the L2,1-norm (NMF-L2,1) was developed.
• L2,1-norm minimization was applied to both the error function and regularization term.
• The method was tested on plant and tumor gene expression datasets.

Main Results:

• The NMF-L2,1 method demonstrates robustness to outliers and noise.
• The approach generates sparse results, aiding in characteristic gene identification.
• NMF-L2,1 successfully extracted more characteristic genes than state-of-the-art methods.

Conclusions:

• NMF-L2,1 offers a unified and effective solution for characteristic gene selection from challenging gene expression data.
• The method's robustness and ability to generate sparse results improve biological discovery.
• This approach advances the field of bioinformatics by enhancing gene selection capabilities.