Interpretable multi-omics integration with UMAP embeddings and density-based clustering.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Integrating multi-omics data is essential for a comprehensive understanding of biological systems.
• Existing methods struggle to capture complex, non-linear relationships within high-dimensional cellular data.
• Single omics approaches often provide incomplete insights into biological control mechanisms.

Purpose of the Study:

• To develop a novel, unsupervised method for integrating diverse omics data types.
• To uncover non-linear relationships between genes, proteins, and metabolites.
• To facilitate interpretable visualization and biomarker identification from integrated multi-omics datasets.

Main Methods:

• Developed GAUDI (Group Aggregation via UMAP Data Integration), a non-linear, unsupervised integration method.
• Utilized independent UMAP (Uniform Manifold Approximation and Projection) embeddings for concurrent analysis of multiple omics data.
• Applied the method to cluster samples based on multi-omic profiles and identify latent factors within each omics layer.

Main Results:

• GAUDI demonstrated superior performance in uncovering non-linear relationships compared to state-of-the-art methods.
• The method successfully clustered samples according to their integrated multi-omic profiles.
• GAUDI identified latent factors, providing interpretable features contributing to sample clusters.

Conclusions:

• GAUDI offers a powerful and interpretable approach for multi-omics data integration.
• The method enhances the identification of novel biological insights and potential biomarkers.
• GAUDI facilitates intuitive visualizations for complex biological data analysis across various experimental designs.