A Self-Training Subspace Clustering Algorithm under Low-Rank Representation for Cancer Classification on Gene...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate cancer classification is crucial for effective diagnosis and treatment.
• Gene expression data offers potential for cancer classification but faces challenges due to high dimensionality and limited sample sizes.
• Existing methods struggle with the complexity of gene expression profiles for reliable cancer identification.

Purpose of the Study:

• To introduce a novel self-training subspace clustering algorithm under low-rank representation (SSC-LRR) for enhanced cancer classification using gene expression data.
• To address the limitations of high-dimensional gene expression data in accurate cancer identification.
• To identify novel gene biomarkers for cancer detection.

Main Methods:

• Low-rank representation (LRR) was employed to extract discriminative features from high-dimensional gene expression data.
• A self-training subspace clustering (SSC) method was utilized for generating cancer classification predictions.
• The proposed SSC-LRR algorithm was evaluated on two benchmark datasets against four state-of-the-art classification methods.

Main Results:

• The SSC-LRR algorithm achieved an overall accuracy of 89.7% and a general correlation of 0.920.
• Performance metrics were significantly higher than the best control method, showing an 18.9% increase in accuracy and a 24.4% increase in correlation.
• Several genes, including RNF114, HLA-DRB5, USP9Y, and PTPN20, were identified as potential novel cancer identifiers.

Conclusions:

• The SSC-LRR algorithm presents a sensitive and effective approach for cancer classification from large-scale gene expression data.
• The identified genes warrant further clinical investigation as potential biomarkers for cancer diagnosis.
• This study highlights a promising new avenue for leveraging gene expression data in cancer research and clinical applications.