Gene2DisCo: Gene to disease using disease commonalities

Area of Science:

• Genomics
• Computational Biology
• Biomedical Informatics

Background:

• Identifying genes associated with complex diseases (disease-genes) is crucial but challenging due to limited known disease-genes.
• Heterogeneous data, often in network formats, and shared disease profiles offer opportunities to improve gene prioritization (GP).

Purpose of the Study:

• To systematically compare disease similarity measures.
• To develop a flexible gene prioritization framework that integrates disease similarities and heterogeneous data, specifically addressing the scarcity of known disease-genes.

Main Methods:

• Developed Gene2DisCo, a novel network-based method using generalized linear models (GLMs).
• Employed an efficient negative selection procedure and imbalance-aware GLMs to handle the scarcity of known disease-genes.
• Designed Gene2DisCo as a flexible framework, independent of specific data types or disease ontologies.

Main Results:

• Gene2DisCo significantly outperformed the benchmark algorithm (kernelized score functions) on a dataset of nine human networks and 708 MeSH diseases.
• Achieved superior performance with an area under the ROC curve of 0.94 compared to 0.86.
• Extended analysis to the whole human genome, identifying top-ranked candidate genes for diseases lacking known annotations.

Conclusions:

• Gene2DisCo provides a robust and flexible framework for gene prioritization, effectively leveraging network data and disease similarities.
• The method demonstrates significant improvements in identifying potential disease-genes, even in data-scarce scenarios.
• Offers valuable candidate gene predictions for unannotated diseases, advancing biomedical research.