BioSTD: A New Tensor Multi-View Framework via Combining Tensor Decomposition and Strong Complementarity Constraint...

Area of Science:

• Computational biology
• Cancer research
• Bioinformatics

Background:

• Omics technologies advance disease mechanism understanding, offering new cancer research avenues.
• Multi-omics data analysis is crucial for cancer research but faces challenges with high dimensionality and integrating global/local features.
• Traditional matrix models struggle to fully leverage the high-dimensional global structure and local information within multi-omics data.

Purpose of the Study:

• To propose a novel tensor integrative framework, the strong complementarity tensor decomposition model (BioSTD), for cancer multi-omics data analysis.
• To identify cancer subtype-specific genes and cluster subtype samples using multi-omics data.
• To address the limitations of traditional methods in capturing high-dimensional global structure and local features.

Main Methods:

• Developed the strong complementarity tensor decomposition model (BioSTD) utilizing multi-view tensors to coordinate omics data.
• Incorporated a strong complementarity constraint to explore local information and enhance subtype separability.
• Applied the model to real cancer datasets for gene identification and sample clustering.

Main Results:

• BioSTD effectively maximizes high-dimensional spatial relationships by coordinating different omics data.
• The strong complementarity constraint enhances the separability of cancer subtypes, capturing consistency and complementarity.
• Experimental results demonstrate that BioSTD outperforms advanced models in analyzing real cancer datasets.

Conclusions:

• BioSTD offers a powerful new approach for comprehensive analysis of cancer multi-omics data.
• The model's ability to integrate global and local features improves the identification of cancer subtypes and specific genes.
• The findings validate the effectiveness of tensor decomposition and the strong complementarity concept in cancer research.