PathIntegrate: Multivariate modelling approaches for pathway-based multi-omics data integration.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• The exponential growth of multi-omics data necessitates advanced integration and interpretation techniques.
• Existing methods often yield complex molecular lists, hindering biological process discovery.
• Interpreting multi-omics data requires significant time and domain expertise.

Purpose of the Study:

• To introduce PathIntegrate, a novel pathway-based method for multi-omics data integration.
• To provide interpretable models by leveraging biological system knowledge.
• To facilitate the analysis of complex, high-dimensional biological datasets.

Main Methods:

• PathIntegrate transforms multi-omics data to the pathway-level using single-sample pathway analysis.
• It employs predictive single-view or multi-view models for data integration.
• Outputs include ranked pathways, omics layer contributions, and molecular importance within pathways.

Main Results:

• PathIntegrate effectively detects signals in low signal-to-noise scenarios by grouping molecules into pathways.
• The method precisely identifies important pathways even at low effect sizes.
• Demonstrated utility in analyzing COPD and COVID-19 multi-omics data.

Conclusions:

• PathIntegrate provides a powerful and interpretable framework for multi-omics data integration.
• The pathway-centric approach simplifies the identification of biologically relevant signals.
• The open-source Python package facilitates broader adoption in biological research.