Disease-gene prediction based on preserving structure network embedding.

Area of Science:

• Computational biology
• Genetics
• Bioinformatics

Background:

• Genetic abnormalities are primary causes of neurodegenerative diseases like Alzheimer's disease (AD) and Parkinson's disease (PD).
• Existing computational methods for predicting pathogenic genes from disease-gene networks have limitations in effectively mining complex relationships.

Purpose of the Study:

• To develop an advanced computational method for predicting disease-associated genes.
• To enhance the accuracy of pathogenic gene identification by integrating diverse biological data.

Main Methods:

• Constructed a heterogeneous network integrating disease-gene, protein-protein, and disease-disease associations.
• Employed a structure-preserving network embedding technique (PSNE) to extract low-dimensional node features.
• Reconstructed a novel disease-gene heterogeneous network using extracted features for improved prediction.

Main Results:

• The PSNE method demonstrated superior performance in disease-gene prediction compared to existing advanced methods.
• Applied PSNE to identify potential pathogenic genes for age-associated diseases, including AD and PD.
• Validated the predicted genes through literature review, confirming their potential relevance.

Conclusions:

• PSNE offers an effective computational approach for disease-gene prediction.
• The study identified high-confidence candidate pathogenic genes for AD and PD, valuable for experimental validation.
• This work contributes to advancing the discovery of disease-related genes through network analysis.