toscca

toscca performs Thresholded Ordered Sparse Canonical Correlation Analysis (TOSCCA) for integrating multiple high-dimensional datasets from omics and imaging devices to discover and understand underlying biological mechanisms.

Key Features:

• Sparse CCA via soft-thresholding: Produces near-orthogonal components without tuning a penalty parameter.
• Sparsity browsing: Allows exploration over various sparsity levels.
• Permutation-based hypothesis testing: Provides statistical significance assessment for discovered correlations.
• Reduced initialization dependence: Results are less dependent on initialization compared to alternative approaches.
• Improved interpretability: Offers enhanced interpretability compared to Penalized Matrix Analysis (PMA) without compromising signal discovery.

Scientific Applications:

• Cancer genomics: Illustrated on real cancer genomics data from drug sensitivity screens.
• Omics data integration: Integrates two or more high-dimensional omics datasets for biological mechanism discovery.
• Imaging-omics integration: Combines imaging and omics data collected in clinical and biomedical research.

Methodology:

Soft-thresholding approach for sparse canonical correlation analysis with near-orthogonal components; performance examined with simulations and compared to Penalized Matrix Analysis (PMA).