Predicting candidate genes from phenotypes, functions and anatomical site of expression

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning

Background:

• Current disease-gene prioritization methods rely on incomplete phenotype-gene associations.
• Gene function and expression data offer a more comprehensive information source.

Purpose of the Study:

• To develop a novel graph-based machine learning method for biomedical ontologies.
• To improve gene-disease association prediction by integrating diverse biological data.

Main Methods:

• Developed a graph-based machine learning approach to embed genes using phenotypes, functions, and expression data.
• Utilized biomedical ontologies and graph-structured data.
• Built a machine learning model to predict gene-disease associations.

Main Results:

• The developed method significantly improves gene-disease association prediction over state-of-the-art approaches.
• Extended phenotype-based gene prioritization to include genes with known functions or expression sites.
• Successfully embedded genes based on phenotypes, functions, and anatomical expression locations.

Conclusions:

• The novel machine learning method enhances gene prioritization by leveraging comprehensive biological information.
• This approach expands the applicability of gene prioritization beyond known phenotype-gene links.
• The method offers a significant advancement in identifying potential gene-disease relationships.