A multiobjective multi-view cluster ensemble technique: Application in patient subclassification.

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• High-throughput omics technologies generate large biomedical datasets.
• Clustering single omics data is valuable, but integrating multiple omics data types is crucial for comprehensive analysis.
• Patient sub-classification requires a combined analysis of all available omics data.

Purpose of the Study:

• To propose a late integration based multiobjective multi-view clustering algorithm for improved patient sub-classification.
• To effectively combine and analyze diverse omics datasets for deeper biological insights.

Main Methods:

• Generated diverse clustering solutions (base partitionings) for each omic dataset using k-means, complete linkage, spectral, and fast search clustering.
• Combined base partitionings using a novel perturbation operator with an ensemble technique.
• Optimized objective functions (conn-XB and agreement index) using multiobjective simulated annealing (AMOSA).
• Combined non-dominated solutions based on similarity to generate a single set.

Main Results:

• The proposed algorithm was evaluated on 13 multi-view cancer datasets.
• Demonstrated effectiveness through an elaborated comparative study against baseline and state-of-the-art methods.

Conclusions:

• The developed algorithm offers a powerful approach for multi-omics data integration and clustering.
• This method has the potential to reveal deep insights for patient sub-classification and biomedical research.