Network-Based Approaches for Disease-Gene Association Prediction Using Protein-Protein Interaction Networks

Area of Science:

• Computational biology
• Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) identify disease-related genomic regions but are resource-intensive.
• Network-based computational approaches offer efficient disease-gene association prediction.
• These methods assume disease-related genes cluster within molecular networks like protein-protein interaction (PPI) networks.

Purpose of the Study:

• To survey network-based disease-gene association prediction methods.
• To categorize these methods into graph-theoretic, machine learning, and integrated approaches.
• To compare the performance of selected methods using a heterogeneous network.

Main Methods:

• Constructed a heterogeneous network integrating PPI, disease, and known disease-gene association data.
• Evaluated six network-based prediction methods.
• Compared performance in scenarios with and without prior knowledge of disease-associated genes.

Main Results:

• The integrative method HerGePred excelled when known disease genes were available.
• The network propagation algorithm PRINCE performed best when known disease genes were absent.
• Integrative and heterogeneous network approaches generally outperformed others.

Conclusions:

• Network-based methods are valuable for disease-gene association prediction.
• Integrative methods and heterogeneous networks enhance prediction accuracy.
• Method selection may depend on the availability of prior genetic data.