SMCC: A Novel Clustering Method for Single- and Multi-Omics Data Based on Co-Regularized Network Fusion.

Area of Science:

• Computational biology
• Bioinformatics
• Data science

Background:

• Clustering is vital for statistical analysis and precision medicine.
• Integrating multi-omics data for cancer subtype identification is challenging.
• Existing network fusion models often fail to maintain data distribution consistency.

Purpose of the Study:

• To develop a flexible network fusion clustering model that minimizes distribution differences.
• To improve data fusion performance by addressing inconsistencies between networks.
• To create a model applicable to both single- and multi-omics data for subtype discovery.

Main Methods:

• Proposing a novel co-regularized network fusion model (SMCC).
• Integrating low-rank subspace representation and entropy for network fusion.
• Measuring and minimizing distribution differences between similarity and fusion networks via co-regularization.

Main Results:

• SMCC reduces noise interference and enhances statistical consistency of fused data.
• Evaluated on 16 real-world datasets, SMCC outperformed 17 state-of-the-art methods.
• Demonstrated superior clustering performance in identifying cancer and cell subtypes.

Conclusions:

• SMCC offers a robust approach for single- and multi-omics data clustering.
• The model's ability to preserve data distribution is key to its effectiveness.
• SMCC significantly contributes to advancing precision medicine through improved subtype identification.