PGAGP: Predicting pathogenic genes based on adaptive network embedding algorithm.

Area of Science:

• Computational biology
• Bioinformatics
• Genomics

Background:

• Disease-gene association is crucial for understanding human health.
• Extracting accurate pathogenic gene information from vast biomedical data is challenging.
• Existing computational methods require improvement for speed and precision.

Purpose of the Study:

• To introduce PGAGP, a novel computational method for pathogenic gene prediction.
• To enhance disease-gene network analysis using adaptive network embedding.
• To improve the accuracy and efficiency of identifying disease-related genes.

Main Methods:

• Utilized Gaussian random projection for initial node feature extraction.
• Employed an adaptive refining process to optimize node features.
• Applied network propagation to an improved heterogeneous disease-gene network.

Main Results:

• PGAGP demonstrated superior predictive performance compared to state-of-the-art algorithms.
• Experiments validated the effectiveness of PGAGP's parameters and integration strategies.
• Case studies confirmed predicted genes through literature verification and enrichment analysis.

Conclusions:

• PGAGP offers an effective solution for pathogenic gene prediction.
• The method enhances the mining of heterogeneous disease-gene networks.
• This approach advances computational strategies in disease-gene association studies.