Gene selection for microarray data classification via adaptive hypergraph embedded dictionary learning.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Microarray data classification is crucial for cancer diagnosis but challenging due to high dimensionality (many genes, few samples).
• Dimensionality reduction is essential for effective gene expression data analysis.
• Existing methods struggle to capture complex biological data structures.

Purpose of the Study:

• To develop a novel computational gene selection model for enhanced microarray data classification.
• To address the challenge of high-dimensional gene expression data in cancer research.
• To improve the accuracy of cancer diagnosis, treatment, and prevention through better data analysis.

Main Methods:

• Introduced adaptive hypergraph embedded dictionary learning (AHEDL) for gene selection.
• Learned a dictionary from high-dimensional microarray data and used reconstruction coefficients for gene representation.
• Employed l2,1-norm regularization for row sparsity to select discriminative genes.
• Adaptively learned and embedded a hypergraph to capture local manifold structures in high-order.
• Developed an iterative updating algorithm to solve the optimization problem.

Main Results:

• The proposed AHEDL model demonstrated superior performance in microarray data classification.
• Experiments on six public microarray datasets confirmed the efficacy of AHEDL.
• AHEDL outperformed other state-of-the-art methods in classification accuracy.

Conclusions:

• AHEDL is an effective computational method for gene selection in high-dimensional microarray data.
• The model successfully captures complex data structures and improves classification outcomes.
• AHEDL offers a promising approach for advancing cancer diagnosis and related research.