Random walk with restart on multilayer networks: from node prioritisation to supervised link prediction and beyond

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Biological networks are crucial for knowledge representation.
• Multilayer networks integrate diverse, multi-scale data.
• MultiXrank is a Random Walk with Restart algorithm for multilayer networks.

Purpose of the Study:

• To demonstrate the versatility of MultiXrank in various bioinformatics tasks.
• To showcase the application of MultiXrank in prioritizing biological entities and predicting associations.
• To illustrate the use of MultiXrank for disease signature identification.

Main Methods:

• Utilized MultiXrank to explore multilayer networks of genes, drugs, and diseases.
• Employed MultiXrank scores in a supervised learning framework for a binary classifier.
• Computed diffusion profiles using MultiXrank on a multilayer network with genomic information.

Main Results:

• Prioritized genes and drugs by exploring interaction networks.
• Successfully trained a classifier to predict gene-disease associations, validated with novel data.
• Identified shared phenotypic characteristics of immune diseases through diffusion profile clustering.

Conclusions:

• MultiXrank demonstrates broad applicability in bioinformatics.
• The algorithm's versatility supports diverse computational biology tasks.
• Further bioinformatics applications are anticipated.