Multiview clustering of multi-omics data integration by using a penalty model.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Multiview clustering and multi-omics data integration methods face challenges with data noise and limited sample sizes.
• Current integration techniques struggle with consistent and differential cluster patterns, leading to limited performance and accuracy.

Purpose of the Study:

• To develop a computational framework for multiview clustering based on a penalty model to enhance multi-omics data integration.
• To address limitations in accuracy and performance when integrating data with consistent and differential cluster patterns.

Main Methods:

• A novel multiview clustering computational framework utilizing a penalty model was developed.
• The method's performance was assessed using synthetic datasets and four real-world multi-omics datasets.
• Evaluations included comparisons with existing literature approaches under various scenarios.

Main Results:

• The proposed method demonstrated competitive performance against current techniques for consistent clusters in synthetic data.
• Enhanced performance was observed for differential clusters.
• The developed method showed superior performance on real omics data, providing richer information and better integration of consistent and differential cluster patterns.
• The method identified significant differences in survival times across various cancer types.

Conclusions:

• A new multiview clustering method was successfully developed and validated using both synthetic and real data.
• This method outperforms existing techniques in integrating multi-omics data with consistent and differential cluster patterns.
• The approach effectively determines significant differences in survival times.