Robust integrative biclustering for multi-view data.

Area of Science:

• Biomedical data analysis
• Computational biology
• Bioinformatics

Background:

• Biomedical research often involves analyzing multiple data views (e.g., genomics, proteomics).
• Detecting sample subgroups linked to specific variable groups is crucial for understanding complex diseases.
• Existing multi-view biclustering methods often require complex parameter tuning, limiting practical application.

Purpose of the Study:

• To extend a sparse singular value decomposition biclustering method for multi-view data.
• To develop an integrative sparse singular value decomposition (iSSVD) algorithm that is user-friendly and computationally efficient.
• To improve the detection of stable, interpretable biclusters in multi-view biomedical datasets.

Main Methods:

• Extended a sparse singular value decomposition (SVD) approach for single-view data to handle multiple data views.
• Incorporated stability selection to control Type I error rates and estimate bicluster membership probabilities.
• Developed the integrative sparse singular value decomposition (iSSVD) algorithm.

Main Results:

• iSSVD successfully identified stable biclusters with interpretable sample-variable associations.
• Simulations and real-world data analyses demonstrated iSSVD's superior performance compared to existing methods.
• The algorithm effectively detects meaningful biological patterns in multi-view data.

Conclusions:

• Integrative sparse singular value decomposition (iSSVD) offers a robust and efficient solution for multi-view biclustering.
• iSSVD is a valuable tool for disease subtyping and uncovering complex relationships in biomedical data.
• The method's stability selection and probability estimation enhance the reliability and interpretability of bicluster findings.