msPLS

msPLS models multiset high-dimensional omics data to identify pathway-level associations and sparse biomarkers that explain phenotypic variation.

Key Features:

• Simultaneous Modeling Across Omics Domains: msPLS concurrently models multiple molecular markers from different omics domains (e.g., genomics, proteomics) to assess their effects on phenotypic variation.
• Hierarchical Structure Consideration: msPLS accounts for the inherent hierarchical relationships between different omics data sources during modeling.
• Sparse Solutions for Interpretability: msPLS employs sparsity-inducing penalties to produce interpretable results from hundreds of thousands of biomolecular variables.
• Pathway Exploration and Biomarker Identification: Through simulation studies and real-world applications, msPLS discovers associated variables and explores biological pathways linked to complex phenotypes such as Marfan syndrome and Chronic Lymphocytic Leukaemia (CLL).
• Comparative Performance: In comparative analyses with Multi-Omics Factor Analysis (MOFA), msPLS demonstrated superior performance in explaining variation, exemplified by identifying two critical clinical markers in a CLL dataset.

Scientific Applications:

• Pathophysiological Insights: By analyzing high-dimensional omics datasets, msPLS provides insights into biological pathways associated with diseases such as Marfan syndrome and CLL.
• Biomarker Discovery: msPLS identifies significant biomarkers to support diagnostic and therapeutic research and personalized medicine approaches.

Methodology:

msPLS uses a multiset multivariate approach with sparsity-inducing penalties and accounts for hierarchical relationships between omics sources to integrate multiple omics datasets into interpretable models.