Multi-layer matrix factorization for cancer subtyping using full and partial multi-omics dataset.

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Cancer heterogeneity necessitates accurate subtyping for effective treatment.
• Current multi-omics approaches often fail with incomplete datasets.
• Implicit relationships across omics data layers are underexplored.

Purpose of the Study:

• To develop a novel computational approach for cancer subtyping using multi-omics data.
• To address the challenge of missing data in multi-omics cancer datasets.
• To improve the accuracy and robustness of cancer subtype identification.

Main Methods:

• Introduced Multi-Layer Matrix Factorization (MLMF) for multi-omics data clustering.
• Applied multi-layer factorization to derive latent feature representations for each omics type.
• Fused latent representations and employed spectral clustering for subtype determination.
• Incorporated a class indicator matrix to manage missing omics data.

Main Results:

• MLMF demonstrated comparable or superior performance to state-of-the-art methods on 12 cancer datasets.
• The approach effectively handled both complete and incomplete multi-omics data.
• Validated the ability of MLMF to capture implicit relationships across omics layers.

Conclusions:

• MLMF provides a unified framework for cancer subtyping with complete or incomplete multi-omics data.
• The method enhances subtype discovery by integrating information across multiple omics layers.
• MLMF offers a robust and open-source solution for cancer research.