HerGePred: Heterogeneous Network Embedding Representation for Disease Gene Prediction.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Identifying disease-causing genes is crucial for understanding diseases and developing cures.
• Existing computational methods for disease gene prediction often struggle to fully leverage heterogeneous data, including disease symptoms and gene-related information like ontology and protein interactions.

Purpose of the Study:

• To develop an advanced computational framework, HerGePred, for enhanced disease gene prediction.
• To improve the accuracy and performance of identifying disease-associated genes by integrating diverse biological data sources.

Main Methods:

• Developed a heterogeneous disease-gene-related network (HDGN) embedding representation framework (HerGePred).
• Obtained low-dimensional vector representations (LVR) of network nodes.
• Proposed two prediction algorithms: LVR-based similarity prediction and random walk with restart on a reconstructed network (RW-RDGN).

Main Results:

• The LVR effectively preserves both local and global network structures within the HDGN.
• RW-RDGN demonstrated superior performance compared to state-of-the-art algorithms in tenfold cross-validation and external validation.
• The framework successfully predicted disease candidate genes, aiding molecular mechanism investigation.

Conclusions:

• HerGePred provides a powerful and effective approach for disease gene prediction.
• The RW-RDGN algorithm represents a significant advancement in computational disease gene identification.
• Accurate disease gene prediction is vital for advancing biomedical research and therapeutic development.