Heterogeneous data integration methods for patient similarity networks.

Area of Science:

• Biomedical Informatics
• Machine Learning
• Clinical Research

Background:

• Patient similarity networks (PSNs) represent patient relationships using nodes and weighted edges.
• PSNs aid in predicting patient outcomes, phenotypes, and disease risks using machine learning.
• Visualization of PSNs offers insights into complex patient data and model explainability.

Purpose of the Study:

• To review existing methods for integrating multiple biomedical data views to construct PSNs.
• To survey patient similarity measures applicable to PSN construction.
• To identify machine learning methods not yet applied to PSNs but relevant for heterogeneous data integration.

Main Methods:

• Review of current literature on PSN construction and data integration techniques.
• Analysis of various patient similarity measures.
• Exploration of machine learning algorithms for heterogeneous data fusion.

Main Results:

• Identification of diverse methods for building PSNs from multiple data sources.
• Compilation of patient similarity metrics.
• Highlighting of machine learning techniques applicable to PSN enhancement.

Conclusions:

• Data fusion techniques are crucial for leveraging high-dimensional, heterogeneous patient data in PSNs.
• The review provides a resource for researchers applying machine learning to PSNs.
• Future work should focus on integrating multi-omics, clinical, and imaging data for robust PSN construction.