HTRPCA: Hypergraph Regularized Tensor Robust Principal Component Analysis for Sample Clustering in Tumor Omics Data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer genomics data analysis is crucial but challenged by high dimensionality and noise.
• Traditional matrix-based clustering methods struggle to capture complex geometric structures.
• Existing techniques often fail to effectively handle outliers in omics data.

Purpose of the Study:

• To develop a novel method for cancer genomics data clustering that addresses dimensionality and noise.
• To improve the mining of underlying geometric structures within cancer omics data.
• To enhance the accuracy and robustness of cancer sample clustering.

Main Methods:

• Representing cancer omics data using tensors to manage high dimensionality.
• Applying hypergraph regularization to preserve geometric structure information.
• Utilizing robust principal component analysis for noise and outlier reduction.
• Decomposing data into low-rank and sparse components for effective clustering.

Main Results:

• The proposed hypergraph regularized tensor robust principal component analysis (HTRPCA) model effectively clusters cancer samples.
• HTRPCA successfully retains complex geometric information among samples.
• Experimental results on TCGA datasets show HTRPCA outperforms advanced clustering methods.
• The method isolates underlying sample structures from sparse interference points.

Conclusions:

• HTRPCA offers a robust and effective approach for cancer genomics data clustering.
• The tensor and hypergraph framework enhances the ability to uncover hidden patterns in complex biological data.
• This method provides a significant advancement for cancer subtyping and biomarker discovery.