Multi-omics clustering for cancer subtyping based on latent subspace learning.

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• High-throughput technologies enable multi-omics data analysis for understanding disease etiology.
• Existing computational methods for multi-omics clustering often struggle with incomplete datasets (partial multi-omics).
• Accurate cancer subtype identification is crucial for understanding disease mechanisms and developing targeted therapies.

Purpose of the Study:

• To develop a novel computational method for clustering partial multi-omics data.
• To address the challenge of missing data in multi-omics analyses for cancer subtype identification.
• To improve the efficiency and effectiveness of cancer subtype discovery using multi-omics data.

Main Methods:

• Proposed a novel multi-omics clustering method based on latent subspace learning (MCLS).
• Utilized Principal Component Analysis (PCA) and Singular Value Decomposition (SVD) to construct a latent subspace from complete omics data.
• Projected incomplete multi-omics data into the latent subspace and applied spectral clustering for sample grouping.

Main Results:

• MCLS demonstrated superior efficiency and effectiveness in cancer subtype identification across seven cancer datasets.
• The method successfully handled partial multi-omics data, outperforming state-of-the-art approaches.
• Experimental results validated MCLS's capability in analyzing multi-omics data with missing values.

Conclusions:

• The proposed MCLS method provides a robust solution for clustering partial multi-omics data in cancer research.
• MCLS facilitates a more comprehensive understanding of cancer and its underlying biological mechanisms.
• The method offers valuable insights for identifying distinct cancer subtypes, aiding in personalized medicine approaches.