KDGene: knowledge graph completion for disease gene prediction using interactional tensor decomposition.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Accurate identification of disease-associated genes is vital for understanding disease mechanisms.
• Current methods using biological networks and deep learning often neglect complex relationships in biological knowledge graphs.
• Existing knowledge graph embedding techniques show suboptimal performance on domain-specific biological data.

Purpose of the Study:

• To develop an advanced framework for disease gene prediction by incorporating complex relations from biological knowledge graphs.
• To improve the representation of semantically similar biological concepts and enhance disease gene identification accuracy.
• To provide a scalable solution for identifying novel candidate disease genes.

Main Methods:

• Construction of a biological knowledge graph focused on diseases and genes.
• Development of KDGene, an end-to-end knowledge graph completion framework utilizing interactional tensor decomposition.
• Incorporation of an interaction module to bridge entity and relation embeddings within tensor decomposition.

Main Results:

• KDGene significantly outperforms existing state-of-the-art disease gene prediction and general knowledge graph embedding methods.
• Comprehensive biological analysis validates KDGene's capability in accurately identifying novel candidate disease genes.
• The proposed framework demonstrates scalability and promising results for future experimental validation.

Conclusions:

• KDGene offers a superior approach to disease gene prediction by effectively utilizing biological knowledge graph information.
• The framework's ability to capture complex relationships enhances the accuracy and discovery of disease-associated genes.
• This work provides a valuable resource for researchers seeking to identify candidate genes for further wet experiments.