ViLoN-a multi-layer network approach to data integration demonstrated for patient stratification.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Modern network representations leverage diverse data sources and patient similarities.
• Patient stratification is crucial for clinical practice and understanding disease mechanisms.
• Integrating multiple molecular profiles presents a significant challenge.

Purpose of the Study:

• Introduce a novel network-based approach, Variation of Information fused Layers of Networks (ViLoN), for multi-omics data integration.
• To enhance patient stratification by directly incorporating prior functional knowledge (KEGG, GO).
• To validate ViLoN's performance across various molecular data types and cohort sizes.

Main Methods:

• Developed the ViLoN algorithm for network construction and patient representation using pathway networks.
• Integrated multiple molecular profiles including gene expression, methylation, and copy number data.
• Incorporated prior functional knowledge from KEGG and Gene Ontology (GO) databases.

Main Results:

• ViLoN demonstrated substantial improvements and consistently competitive performance in patient stratification across multiple data type combinations.
• The integration of prior functional knowledge was critical for achieving good results, particularly in smaller patient cohorts.
• ViLoN outperformed alternative methods in smaller cohorts (e.g., 90 patients for rectum adenocarcinoma, 180 for esophageal carcinoma).

Conclusions:

• ViLoN offers a robust and effective method for integrating multi-omics data for patient stratification.
• The incorporation of functional knowledge significantly enhances the performance of network-based patient stratification, especially in data-limited scenarios.
• ViLoN represents a valuable advancement for precision medicine and disease research.