Multi-cancer samples clustering via graph regularized low-rank representation method under sparse and symmetric...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer type identification from gene expression data is a key bioinformatics challenge.
• High-dimensional, noisy gene expression data complicates accurate cancer sample clustering.
• Existing clustering methods may have limitations in handling complex genomic data.

Purpose of the Study:

• To develop an improved method for clustering multi-cancer samples based on gene expression data.
• To address the challenges posed by data noise and high dimensionality in cancer genomics.
• To enhance the accuracy and robustness of cancer subtyping.

Main Methods:

• Proposed a novel graph regularized low-rank representation under symmetric and sparse constraints (sgLRR) method.
• Integrated manifold learning-based graph regularization and symmetric sparse constraints into traditional low-rank representation (LRR).
• Applied sgLRR to decompose gene expression data, followed by spectral clustering using normalized cuts (Ncuts) on an affinity matrix.

Main Results:

• The sgLRR method effectively alleviates noise effects through symmetric and sparse constraints.
• Graph regularization preserves the intrinsic local geometrical structures of the gene expression data.
• Clustering multi-cancer samples using sgLRR demonstrated improved clustering quality compared to existing methods.

Conclusions:

• The proposed sgLRR method offers significant advantages for multi-cancer sample clustering.
• Experimental results confirm the remarkable performance of sgLRR in handling gene expression data.
• This approach enhances the potential for accurate cancer subtyping and personalized medicine.