Gene gravity-like algorithm for disease gene prediction based on phenotype-specific network.

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Polygenic diseases arise from multiple gene dysfunctions, necessitating advanced methods for disease gene discovery.
• Network-based approaches utilizing 'omic' data have advanced disease gene identification, yet integrating diverse data types remains a challenge.

Purpose of the Study:

• To enhance disease gene prediction by integrating phenotype similarity, biological function, and network topology.
• To develop and validate a novel network construction and gene scoring algorithm for improved polygenic disease gene discovery.

Main Methods:

• Constructed phenotype-specific networks by mapping phenotype similarity onto protein-protein interaction (PPI) networks.
• Developed a gene gravity-like algorithm integrating topological and functional similarity for gene scoring.
• Validated the approach using leave-one-out and leave-10%-out cross-validation against state-of-the-art algorithms and DisGeNET database.

Main Results:

• Phenotype-specific networks and the gene gravity-like algorithm demonstrated superior performance in disease gene prediction.
• Cross-validation confirmed the effectiveness of the proposed methods compared to existing algorithms.
• Predicted disease genes for obesity, prostate cancer, and lung cancer showed high consistency with existing literature and database evidence.

Conclusions:

• Phenotype similarity information significantly enhances disease gene prediction accuracy.
• The gene gravity-like algorithm, combining topological and functional similarities, outperforms Random Walk with Restart (RWR).
• This study provides valuable insights into disease gene discovery through the fusion of multi-similarity data.